The AI Everywhere book series brings together multidisciplinary voices to explore how artificial intelligence is being developed, who is shaping it, and what narratives are often left out. Rather than treating AI as purely technical innovation, it emphasizes history, responsibility, and the human decisions behind the systems we use every day.

Drawing from her background as an embodied philosopher with degrees in neuroscience, and her work as the host of , Andrea explores how AI can be oriented toward meaningful human and ecological connection. She reframes AI as a vehicle that humans must consciously orient toward a purpose, rather than focusing solely on the “loop” between a technology and a person.

In this conversation with , Andrea’s insights uncover why we must stop “mistaking the map for the territory” when interacting with large language models. She explores how to break the “trance” of automated interaction and why preserving relational friction, the tension and friction that define real life, is the most powerful way to ensure that technology serves our collective well-being.

To join future live video interviews and events, subscribe to SheWritesAI and get the Substack app:

Subscribe now

Highlights from the Live Interview

The Global Tribe & The Philosophy of Care

Question: Why was it important for you to position Care as a central theme in AI Everywhere?

Andrea: Well, I think it’s incredibly important. It’s actually probably the most important thing we could think about right now, but nobody’s really thinking about it because we are caught in a loop.

We have a new tool and we can think of it as like a vehicle, but we have to orient it. And right now, we are not orienting it towards care and connection with humans — relational, ecological connection.

Instead, we’re kind of worrying about this loop between technology and person. Care is like the movement that we are making through life. So what we care about is actually orienting us.

Care is power. And that’s what you’re giving to those machines. Care is actually a relational friction. Equilibrium is actually death. When you get into a state where you’re completely smooth, then you are no longer alive. The tension and the friction is what life is.

Don’t think of AI as the end. Think of it as the vehicle that you’re going to orient towards something in the real world.

Question: Beyond Either/Or: How does moving past binary thinking allow us to better use these tools?

Andrea: I think it’s important to understand it’s not neutral, of course, but that the technology isn’t neutral doesn’t mean that it’s only either one thing or the other. And the more we try to do that, the more we sort of close off from one another.

And we call it like a savior or a monster; if we try to talk about them against one another, like one side’s going to win, then we have trouble. I mean, we’re not actually getting anywhere.

Again, we’re focusing on that loop instead of like looking at the landscape and thinking, where do we want to orient this?

Technical Deep Dive: Language as a Bridge

Question: The Mirror of Language: How does interacting with a machine help a human become more aware of their own communication habits?

Andrea: I think what’s coming to me right now is that the idea that we are mistaking the map for the territory, because language is so important to us and it cues so many old habits from our whole life.

So when we’re interacting with language ... we’re actually kind of creating a world that’s based on all other interactions we’ve ever had, because the language represents that.

And so it’s queuing certain habitual bodily responses that were actually built through relational environmental actions. But we confuse that. It’s happening in that moment, but actually, everything is cumulative. So we’re bringing our whole life to this.

Question: Care as an Opportunity: Can you share a practical example of how a 'transactional' AI interaction can be transformed into a 'relational' one?

Andrea: Noticing when it’s a transaction and letting it be a transaction, but not mistaking it for a relationality.

And then you can sort of use the chatbot to practice. It’s like the car. It’s the vehicle that can help you get to those places. But if you’re not thinking of it like that, then you’re just getting sucked into it and it’s just habituating you thinking that is the relationship.

Maybe I can say, “Okay, I’m going to have like an experiment with an LLM.” It’s practice. I’m not going to ask the LLM these questions. I’m instead going to ask the LLM to help me express myself. Once you start getting able to say those kinds of things, you’ve started that habit. And then it only makes sense though, if you then go to the landscape and the environment and the people around you.

The Interconnected AI Landscape

Question: The "i-risk" of Emptiness: If an AI can simulate 'care' or 'love' perfectly, do we risk losing the authentic 'human signal'?

Andrea: And so what do we care about? Do we care about bodies being alive, relational, being able to not judge each other so harshly, open spaces where we can explore all these amazing possibilities of life that we hardly even see right now?

So I don’t think it’s going to actually replace us unless we are just exhausted and we narrow ourselves into that equilibrium state, which could happen because there is so much overwhelm.

But I really believe in us as humans. I think you love your grandmother and so do I. I love my dog. We love our kids. We love our families. These things are hard. But it’s where the meaning is.

Question: Relational Ethics vs. Wonder Bread: How does your concept of 'Care Beyond Technology' help us move toward relational ethics?

Andrea: Then that relationality becomes the most important thing, not the technology.

The technology becomes the way we can be in our bodies better, and in the world better, and with more excitement, more motivation.

We can’t all be experts at all of this stuff, but we can help each other do that. And that community in that way becomes much more important, real interaction, able to know the people you’re working with in some way and trust them.

Governance & The Accountability of Connection

Question: The Accountability Gap: Who is accountable if the language model subtly manipulates the user's emotions?

Andrea: I think we have to not be afraid to be vulnerable and to hold tension. And the reason I say that is because we have to show that we care about this subject and that we need each other and that we need help.

And that it’s actually a power. This connects us. We can join in this feeling of tension and vulnerability that is humanity, right? It’s about being able to listen, being there, and trying to create technology that’s going to be resonant, that actually hears what you need to be heard.

Question: Women as Architects of Trust: How does a focus on 'Love' and 'Care' fundamentally change the way we should govern LLMs?

Andrea: Just who really want to do this. Want care to be the foundation of what we're building, the policy we're making, all of that.

But what it really comes down to is just being able to listen, being able to be there and being able to try to create technology that’s going to be resonant, which is it’s going to actually hear what you need to be heard and actually help someone who can address that find you.

Actionable Awareness & Future Vision

Question: Mindful AI Integration: How can 'working people' use a chatbot this week to improve a relationship in their 'analog' life?

Andrea: Maybe I can say, “okay, I’m going to have, like, an experiment with an LLM.” It’s practice. I’m not going to ask the LLM, you know, these questions. I’m instead going to ask the LLM to help me express myself.

And then once you start getting able to say those kinds of things, you’ve started that habit. You know you can do it and you know you have care in you.

And then it only makes sense though, if you then go to landscape and the environment and the people around you and slowly maybe see if you can do some of that and find people who might help do some of that.

Reach out.

Final Message

Question: What is the one thing you want a reader to do differently after reading Chapter 7?

Andrea: I've said it, I mean, it's holding paradox. It's care matters. It's community matters. It's you are something special no matter who you are or where you are.

And don't think of AI as the end. Think of it as the vehicle that you're going to orient towards something in the real world.

Subscribe now

Final Summary: Andrea emphasizes that while AI can mimic human language, it lacks the "relational friction" of real life. She defines true care as the ability to hold tension and stay present in our bodies. Her call to action is to use AI as a tool for practice, but to ultimately "break the trance" and return to the "territory" of real-world human connection.

Love & Philosophy

From the heart. Deeply researched. Mostly unscripted. Exploring philosophical, scientific, technological & poetic spaces beyond either/or bounds. It's reasonable to care.

By Andrea Hiott

Why This Matters

We are living through what describes as a "trance" of automation, where artificial intelligence is increasingly being mistaken for true relationality, often embedded into our daily interactions without a clear understanding of how it manipulates our language, our bodily habits, and our sense of connection

While the digital world offers a smooth, frictionless experience, AI systems have rapidly expanded by replicating human patterns, queuing habitual responses that trick us into "mistaking the map for the territory". This shift has raised urgent questions about the numbing or narrowing of the human experience in a society that often prioritizes automated transactions over the messy, necessary friction of real relationships.

’s perspective highlights that AI is not an end in itself, but a "vehicle" that must be consciously oriented by human agency. In her work, including her newsletter Love and Philosophy, she examines the tension between the "either/or" binary, viewing AI as either a savior or a monster, and instead advocates for a path that moves beyond these boxes toward a deeper commitment to care and connection.

She also emphasizes that these concerns are not abstract or distant. The loss of "relational intelligence" and the outsourcing of our "moral muscle" are already influencing how we parent, how we treat the elderly, and how we support one another through vulnerability. She warns that "equilibrium is death," arguing that the tension and friction found in real-world care are precisely what make us alive.

Rather than focusing only on whether AI is "conscious," this conversation centers the present reality: AI is already a vehicle we are driving, and it is currently reshaping our social and ecological landscapes. The question is no longer how fast the technology can go, but how consciously we choose to "break the trance" and orient it toward the real world

So subscribe, stay connected, and be part of the movement.

Thank you to everyone who tuned into my live video! Join me for my next live video in the app.

Get more from SheWritesAI in the Substack app

Available for iOS and Android

Get the app

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe for free to receive new posts like these interviews, weekly digests, & book news and to support our work.

Subscribe now

Share

Here we are in the She Writes AI interview series, where we talk with the authors behind AI Everywhere, Volume 1 to explore their insights, experiences, and guidance on building AI that is ethical, responsible, and human-centered.

This is the 11th episode in the series, and today we are joined by . Her Chapter 5: AI Cinema: Between Craft, Collapse, and the Possibility of a New Language, explores the emergence of AI as not just a set of tools, but as the foundation of an entirely new cinematic grammar.

Elettra Fiumi is a Swiss-based Italian-American award-winning film director, producer, and educator. Her work moves between documentary filmmaking, archival research, and experimental AI-generated forms, often exploring memory, authorship, and the evolving relationship between humanity and technology. She is the founder of Fiumi Studios and the voice behind AI Cinema, her newsletter documenting the creators, tools, and workflows shaping this emerging medium. She teaches AI for Cinema and creative AI practices internationally at universities, museums, and film festivals, and has created work for platforms including Netflix, BBC, and The New Yorker.

In her practice, AI is not positioned as a shortcut or replacement for filmmaking, but as a structural shift in how stories are developed, iterated, and experienced. Her work asks what happens when cinema is no longer defined solely by capture and editing, but by generation, dialogue, and continuous revision between human intent and machine output.

In this conversation, Elettra reflects on AI as a new cinematic language, the tension between speed and authorship, and why the most important decisions in filmmaking remain stubbornly human. Whether you are a filmmaker, educator, or creative working at the edge of new tools, her insights offer a grounded yet expansive view of what it means to create in a moment of technological transformation.

Below is ’s full interview response to ’s questions for AI Cinema: Between Craft, Collapse, and the Possibility of a New Language from AI Everywhere, Volume 1.

Topic: AI Cinema: Between Craft, Collapse, and the Possibility of a New Language (Chapter 5)

Intended Audience: For filmmakers, artists, media educators, and communication professionals who want to move toward a deeper understanding of meaning, intent, and authorship in an AI-mediated world, focusing on the fundamental "why" of human creation.

Part 1: The Professional Path & The Collective Mission

The Expert Perspective: Elettra, your career has spanned from fact-checking and journalism to award-winning film production and teaching. What was the "spark" that made you realize AI needed to be framed as a new "language" to be learned, rather than just a tool for production efficiency?

Elettra: Every new language begins the moment you realize you’ve been thinking in the old one. For me, that moment came when I started learning about prompting and prompt engineering. I realized that the conversations you have with these machines have their own code, one that’s different from how we typically communicate. But the exciting part was that this code wasn’t entirely foreign. It was about tapping into what I already knew from my past: cinematography, lighting, camera types, lenses. All of those decisions I’d been making instinctively on set for years suddenly needed to be articulated, defined, and communicated to the tools. That act of translating visual intuition into language was thrilling.

The production efficiency side matters too, of course. But that’s really about organizing your thoughts and projects so you can get more done and focus on the creative part. The language piece is something deeper: it’s about world-building, about being able to define the rules of your visual universe and have a real conversation with the machine about them. That’s when I understood this wasn’t just a new tool. It was a new form of creative literacy.

The "Home" Perspective: You describe discovering your love for AI tools during maternity leave, a time you call a "personal and professional rupture." Why was that specific moment of solitude and irregular stretches of time so pivotal for your encounter with generative technology?

Elettra: Learning something as complex as AI requires real, dedicated time. I see it now even when I’m generating for a film or vibe coding: I need uninterrupted stretches where I can go deep. The real work happens there. Maternity leave gave me that: quiet, space, and the safety of a structured learning environment. I was enrolled in a master’s program, so I wasn’t experimenting for the first time in front of a client. That was such a luxury.

There was also something more personal driving it. I knew motherhood was going to keep me closer to home, that the constant travel and juggling of deadlines I’d done before would be much harder, at least in those initial months. And then AI opened this door: the ability to organize my work with real efficiency but also to create stories of the wildest imagination. I had a genuine creative burst after Luna was born. Suddenly I could explore all of these ideas in a doable way, without needing to raise funds for crazy concepts or travel the world to execute them. It was truly empowering.

Part 2: The Systems of Decision & The Creative Dialogue

Shifting the Work: You argue that AI doesn't replace filmmaking but "shifts where the work lives." Which phases of the process do you find AI accelerates most effectively, and why do you believe scriptwriting and editing remain "stubbornly human" tasks?

Elettra: The organizational side of filmmaking is where AI accelerates things most dramatically: analyzing large amounts of data, managing timings, logistics, moving parts. It supports the phases that can be repetitive in production or overwhelmingly complex in coordination. That acceleration is real and meaningful.

But scriptwriting and editing are some of my favorite parts of making a film, and they remain stubbornly human for different reasons. With scriptwriting, there’s something that happens when you’re inventing a story that is deeply personal and yet also about imagination, spontaneity, the unexpected, and whatever you’re most curious about. That’s what makes the writing great. AI hasn’t fully reached those liminal spaces where we live as storytellers, where the story becomes magic.

Why AI tools for Editing haven’t stepped up, honestly, baffles me. I’m surprised the tools haven’t improved more. Tools like CapCut and InVideo have done incredible things, and even Seedance. Great editing is based on a sense of rhythm, an understanding of space and silence, a sophisticated mixing of sound effects, music, voiceover, and image. That sense of rhythm pulls everything together and produces results that feel professional, advanced, high-quality. A great human editor is still unbeatable. So much of the story is built in the edit, even when you’ve already figured everything out from the beginning, even when every shot is clear and planned. In that edit room, you encounter again that spontaneity, that human instinct, that AI editing tools still can’t match. But, of course, the fact that you can generate what you envision and are missing, on the timeline and you can’t do a re-shoot — thank you, AI.

Cinema as a System: You define cinema as a "system of decisions", from rhythm and texture to audience duration. How does AI change the stakes of these decisions when audiences might watch a film differently simply because they know it was generated?

Elettra: Right now, many audiences come to AI-generated film with an approach of refusal before they’ve even been open to watching it. There’s real fear and anger that AI is going to erode culture, diminish the history of cinema, threaten its future. That emotional position shapes the viewing experience before a single frame plays.

At the same time, AI is developing its own visual language, a different set of styles and aesthetics that is still being explored. As the tools evolve, we’re seeing creators produce extremely high-quality films where it’s genuinely difficult to tell they were generated. That gap between AI-generated and traditionally shot imagery will continue to diminish.

And that’s where the question gets interesting, and honestly where I sit with some uncertainty myself. If the story remains human, if the edit remains largely human, and the images are of genuinely high quality, then what exactly makes the experience different for the audience? I think the answer is: less and less. The real question isn’t whether something was generated by a machine. It’s whether someone with a point of view made deliberate decisions about rhythm, texture, and duration. Those decisions are what cinema has always been. The origin of the pixels is secondary.

Dialogue with the Machine: You describe creativity as a dialogue where the “difference” between your prompt and the AI’s output is actually the point. How has this process expanded your own creative practice, perhaps even leading you into genres like science fiction that you hadn’t explored before?

Elettra: Sometimes I have an idea that becomes a testing ground for this exact dialogue. I’ve been developing a concept about a fashion show in the jungle for over a year now, and every time I’m testing a new tool or exploring a workflow, I keep going back to that story. I’m endlessly curious to see the different outputs the same base concept generates, and those experiments are genuinely exciting. They expand my creative practice because they allow me to explore the different visual languages I carry within me, all through the lens of a single story.

But AI has also opened up entirely new questions for me. I’ve found myself exploring ideas about artificial wombs, for instance, questions that would have been completely inaccessible to develop financially through traditional production. These are deeply human questions about the body, about motherhood, about technology’s relationship to life. And now I can explore them in multiple visual directions, testing how the future could look without needing to build a set or raise a budget first. The dialogue with the machine hasn’t just expanded my genres. It’s expanded the kinds of questions I’m able to ask.

Part 3: Case Studies - Radical Landscapes & Mamma Robot

Navigating Abundance: Your documentary Radical Landscapes was a ten-year journey through 16 terabytes of data. If you’d had access to tools like NotebookLM earlier, how might it have reduced the friction and isolation of navigating such a vast personal archive?

Elettra: An archive never gives you all its stories at once. With the sheer volume of interviews I had, covering so many different topics, a tool like NotebookLM could have helped me develop many more stories from that same content, find patterns across history and across people’s voices that I might never have seen with what I had before.

That project was also constantly in fundraising mode, and even though it had many collaborators, every collaboration came with the stress of paying for each person’s time. There was never the luxury of iterating and exploring in peace. And even so, I created so many different drafts of the edit. My poor family kept watching version after version, and they were all so different from each other. I spent enormous amounts of time alone with those transcripts, making decisions by myself, just to be efficient with whatever time I could afford from my collaborators.

From one archive, so many stories are born. With a tool like NotebookLM, it would have been a fascinating study to see how the output might have differed from what I ended up making. And honestly, it reminds me that those possibilities still exist. The archive is still there. The stories I didn’t find the first time are still waiting.

The Logic of Emergence: In contrast to your decade-long documentary, Mamma Robot was born from speed and "fake news." Can you explain how you used AI to discover, rather than execute, this film through loops of image and story generation?

Elettra: Discovery and execution are rarely separate for me. They happen at the same time. Mamma Robot started from a piece of fake news that was bombarding my social media feed and that I became obsessed with. I knew I wanted to explore it, but I had so many different ideas and very little time. With the help of ChatGPT, I was able to refine why I wanted to tell this story and how. Why it mattered to me. And that clarity allowed me to start generating visuals even before the story was fully shaped.

Through that process, I realized something important about how I work: as a learner and a doer, I’m at my best when I’m developing the story and the visual side simultaneously. If I have to write everything first and then shoot, it stalls me. It keeps me from reaching my creative potential. But looping between image and story, letting one inform the other in real time, is an incredibly expansive way of being creative. It lets you discover what’s actually inside of you, rather than trying to plan it all out in advance.

Narrative Registers: In Mamma Robot, you used different AI tools (Runway vs. Google’s Veo) to create different "narrative registers." Why is choosing a specific tool a "narrative decision" rather than just a technical one?

Elettra: Every tool carries its own visual and temporal logic. In Mamma Robot, I deliberately used different AI tools for different parts of the film, and that was a storytelling strategy, not a limitation. The fairy-tale sequences were developed with Runway, which has a painterly, dreamlike quality to its motion. When I needed to build a broader narrative context with a more photorealistic look, I turned to Google’s Veo. The shift in tool produced a shift in texture, movement, and realism that was more detailed, more convincing, and in a subtle way, more unsettling.

Rather than smoothing out these differences, I embraced them. Those contrasts became part of the film’s language, signaling different narrative registers: myth and investigation, intimacy and distance, body and system. When you’re telling a story that moves between the mythical and the real, the texture of the image needs to shift with it. A dream sequence shouldn’t feel the same as a documentary passage. In traditional filmmaking, you’d achieve that through changes in film stock, lens choice, lighting, color grading. In AI filmmaking, the model itself becomes that variable. Runway’s slight surrealism tells the audience they’re in a space of imagination. Veo’s photorealism grounds them in something that feels observed. The audience may not consciously register the technical shift, but they feel the emotional one. That’s what makes it a narrative decision, not a technical one. The tools are not interchangeable. They shape meaning.

Part 4: Governance, Ethics & The Human Signal

The Myth of Speed: You mention that AI communities often hype trailers made in "three hours for $200." Why is this narrative misleading, and how do you defend the idea that clients pay for human judgment rather than "keystrokes"?

Elettra: The decisions you make in three hours are actually based on decades of experience. You know how to prompt, you know how to build the story, you know what a specific lighting choice does to the meaning of the storytelling. That speed isn’t effortless. It’s earned.

And frankly, the three-hour claim is often not even true. If you’re developing a consistent style, then building out the character, the location, the props, the costumes, and refining each individual image before you even enter the video generation phase, or if you’re doing something with real specificity, it takes much longer than three hours. The hype makes it seem like a random, almost accidental process, when in reality you’re making dozens of deliberate choices. Out of twenty, forty — hell, sometimes sixty — outputs, you choose one. And that choice is telling. It reveals your style, your taste, and the depth of experience you bring to the project.

When a client hires me, they’re not paying for the minutes it takes to generate. They’re paying for the years it took to master the tools, know the difference between them, know what to generate, what to keep, and what to discard. That judgment is the work. The keystrokes are just the last step.

Cultural Absence: You shared a story about a student trying to generate a coming-of-age story about a Jamaican boy where the AI failed. How does this "cultural absence" in datasets force filmmakers to confront the limits of representation in AI?

Elettra: A tool can only imagine what it has been taught to remember. When something like that happens in the classroom, it makes you realize how much of our humanity, and the diversity of it, is misrepresented in these systems. These models reflect who built them, who curated the data, and whose stories were considered worth preserving. That realization alone is valuable, but it’s not enough.

Between recognizing something is wrong and complaining about it, there needs to be action, otherwise things don’t change. That action can take many forms. Go out and create those datasets. Contact the tool makers and contribute to their libraries. Write about it. Make a film that shows the absence itself. Use the gap as a subject, not just a frustration.

What excites me is that the missing representation can become the very reason to create more stories about the specific thing that isn’t there. The cultural absence doesn’t just reveal a limit. It can become a prompt to fill it. And that shifts the filmmaker from being a passive user of the technology to someone who actively shapes what it knows.

The Contested Ground: As someone trained in fact-checking, you note that AI models are trained on creative work without consent. How do you navigate the tension of working on this "contested ground" while advocating for better systems of compensation?

Elettra: Transparency is the only honest starting point. As someone trained in fact-checking, I was taught that credibility is cumulative, and so is responsibility. I use these tools knowing they are built on contested ground. The models were trained on vast quantities of creative work, photographs, films, illustrations, music, often without the consent or compensation of those who made it. I don’t have a clean resolution for that, and I’m skeptical of anyone who claims to.

What I try to do is remain transparent about the tension, credit sources and influences whenever possible, and advocate for better systems of compensation and consent as they develop. Pretending the problem doesn’t exist, or that individual creators bear no responsibility for engaging with it, is not a position I can hold.

I also think it’s important to be precise about where the real ethical lines are. Creating an original AI character is no different from animation or fiction. But cloning a real person’s likeness or voice without permission is a rights issue that demands proper frameworks. Lumping everything together doesn’t help anyone. The more precisely we think about these distinctions, the better the systems we’ll build.

Part 5: Actionable Takeaways & Final Message

The Final Message: You conclude that the hardest part of filmmaking was never generating the image, but "knowing what the image should say." If a filmmaker is just starting with AI this week, what is the one "stubbornly human" quality you hope they never outsource to the machine?

Elettra: The why. We can all create beautiful images now. The technology has democratized the how in ways that would have seemed impossible even two years ago. But the question that no tool will ever answer for you is: why are you creating this? What do you want the audience to feel, to question, to carry with them after the screen goes dark?

That comes from imagination, and imagination comes from living. From the things you’ve seen, the people you’ve loved, the questions that keep you up at night. A diffusion model doesn’t have those. You do.

So if you’re starting with AI this week, start with the message. Close your eyes and ask yourself what you want to say and why it matters to you. If you can answer that honestly, the tools will serve you. If you can’t, no amount of generation will save you. The hardest part of filmmaking was never making the image. It was knowing what the image should say. That has always been ours, and I hope it always will be.

The hardest part of filmmaking was never making the image. It was knowing what the image should say.

Cinema is a system of decisions. The tools may change, but the decisions remain human.

Leave a comment

Why This Matters

A central idea emerges clearly across this interview: AI does not simply change how films are made, it changes where creative work actually happens. For , the shift is not about replacing craft, but redistributing it across a new system of decisions, where generation, iteration, and storytelling are deeply intertwined.

From archival documentary work to AI-native experimentation, she traces a practice grounded in both structure and emergence. AI accelerates logistics, expands visual possibility, and enables rapid exploration of ideas that would once have been constrained by budget, time, or production scale. Yet what remains unchanged is the core of cinematic authorship: judgment, rhythm, and narrative intent.

Her perspective resists the idea that speed equals creativity. Instead, she reframes filmmaking as a continuous negotiation between intention and output, where tools evolve, but meaning still depends on human choice. The filmmaker’s role does not disappear in this shift; it becomes more explicit, not less.

At the center of her argument is a simple but defining truth: AI can generate images, worlds, and variations, but it cannot determine significance. It cannot answer the question that gives cinema its direction — what is worth saying, and why now.

In a moment where cinema is being redefined from the inside out, ’s work offers both clarity and tension. It suggests that we are not just adopting new tools, we are witnessing the emergence of a new language. And like all languages, its meaning will depend on who speaks it, and with what intent.

Thank you, Elettra Fiumi, for this expansive and deeply reflective contribution to the conversation. you can find her work here AI Cinema and other AI films can be found here in Escape.ai.

Leave a comment

And if you’re interested in more conversations like this, consider subscribing. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, Volume 1, where 26 women from across five continents explore how artificial intelligence is changing the world we live in, and how we can change it for the better.

Subscribe now

References

I wrote an article about the importance of cinema and AI effect — it might interest you:

Lights On by Farida

The Pattern Nobody Wants to Name

Read more

10 days ago · 17 likes · 8 comments · Farida Khalaf and JHong

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device (free)

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by :

Thanks for reading She Writes AI Community! Subscribe (free) to receive new interviews, weekly digests, & book news and to support our work:

Subscribe now

Share

The AI Everywhere book series brings together multidisciplinary voices to explore how artificial intelligence is being developed, who is shaping it, and what narratives are often left out. Rather than treating AI as purely technical innovation, it emphasizes history, responsibility, and the human decisions behind the systems we use every day.

Drawing from her experience at Microsoft and her work as the author of AI Baby, Celeste explores how AI is being developed in an environment driven by speed, competition, and deep uncertainty. At the same time, she challenges the dominant historical narrative that has overlooked key figures such as Ada Lovelace, Grace Hopper, and Radia Perlman.

In this conversation with , Celeste’s insights uncover how the story of AI has been shaped, how women’s contributions have often been erased from the narrative, and why building accountable AI requires confronting both history and the systems being built today.

To join future live video interviews and events, subscribe to SheWritesAI and get the Substack app:

Subscribe now

Highlights from the Live Interview

The Professional “Why”

Question: What made you realize the male-dominated narrative in AI is a systemic risk?

Celeste:
Certainly there is a disparity that has existed for a long time. It’s still only about 30% women in tech, and when you get to the executive level, it drops to around 12%.

The thing about AI is that this is probably the most seismic shift in technology we’ve ever seen in history.

So when you’re baking bias into the data, whether that’s gender bias or racial bias, once these systems are trained, it’s really locked and loaded.

And if that origin is ignored, it really limits how much women feel a part of this new technology, and ultimately it limits how many younger girls see themselves in the field.

The Global Tribe:

Question: How does it feel to open the book and set the stage for this collective mission?

Celeste:
It was really shocking to me to discover so many women I had never heard of who really should be household names. These women have become my heroes.

It was also just really powerful to find a community of women who are deeply concerned about the ethics of AI. That sense of shared concern is really what has drawn me further into this space.

And when you think about these women going into technology in the very early days of computing, it was something truly remarkable.

Ada Lovelace & AI Origination

Question: Does Ada Lovelace’s idea that machines cannot originate still hold today?

Celeste:
That debate is still just as relevant today as it was when she was doing her research.

You would think, wow, AI is actually originating these thoughts but in reality, AI is being fed a massive amount of data by humans.

So if we define origination as creativity, lived experience, and truly original thought, then I would say Ada Lovelace’s theory still holds true.

Women Who Built Computing Foundations

Question: Why is it important to remember women like Dorothy Vaughan, Grace Hopper, and Radia Perlman?

Celeste:
These women have become my heroes. You have to think about these women as the original, the OGs of tech.

At a time when there were so few of them, the outsized contributions they made in laying the foundations of modern computing are profound.

What’s fascinating is that they didn’t just learn the systems — they brought their teams along too. There’s so much to be learned from their perseverance and leadership.

Fei-Fei Li & Data Quality

Question: How did Fei-Fei Li change AI through data?

Celeste:
Fei-Fei Li realized that if you could get enough high-quality data, you could completely change the game for AI. Her work with large, curated datasets set the stage for what we now know as modern AI.

Before her contributions, we were in what was called the ‘AI winter,’ a period of stagnation in the field. But her breakthrough showed that it wasn’t just about having more data — it was about having quality data.

Now, large tech firms are just wholesale scraping data from the internet, and as the saying goes, ‘garbage in, garbage out’.

Scaling Bias

Question: How does AI amplify societal bias?

Celeste:
AI amplifies societal bias because the data it’s trained on is inherently biased. And as you scale those systems to different languages and contexts, the bias just gets magnified. There’s really only a handful of tech conglomerates that have the resources to develop this technology, and they control a lot of the narratives.

If you’re able to anthropomorphize AI, treating it like a human entity, you end up taking responsibility away from the developers and the systems themselves. Ultimately, it’s the algorithms, the data, the weights, and the people behind them that shape how bias gets baked into AI.

Present Harm vs Existential Risk

Question: Why is the extinction narrative dominating instead of present harms?

Celeste:
The focus on the extinction narrative allows people to shift responsibility away from themselves. If you’re constantly looking at the future, at this existential threat of AI, you’re not dealing with the immediate, present harms that are already happening. It’s a way to avoid responsibility for what’s going on right now.

It’s also a bit of a marketing ploy for those pushing the technology, as it draws attention away from the ethical concerns and biases embedded in current AI systems. The women raising these red flags about the harms we’re already facing are often portrayed as slowing down progress, which is simply not true. We need a diverse group of voices to shape this conversation, not just the ones pushing the future hype.

Ethical AI & Industry Resistance

Question: How can companies support ethical AI instead of silencing it?

Celeste:
There’s a clear example with a researcher who wrote a paper exposing biases in AI, and when she did, Google, who was signing her paychecks, didn’t like what she said, so they terminated her.

This story clearly shows that ethical concerns aren’t always welcomed, and it’s really not happening the way it should right now. Companies often face an inherent conflict of interest: they want to push AI forward quickly, but they don’t always prioritize ethics or accountability.

It would be nice if there were a way to address these issues from within the companies, but if ethical considerations aren’t baked into the process from the start, it’s going to be too late to fix things down the line. Building ethical AI has to be part of the foundation, not an afterthought.

AI BABY & Storytelling

Question: How does storytelling help explain AI?

Celeste:
What’s good about fiction is that it really shows the humanity behind the technology. It makes us think about how people will truly be affected by AI, which is permeating everything, and the very human decisions that will need to be made. Fiction allows us to ask, “What would I do in this situation?” It brings the stakes of AI to life in a way that makes it more approachable and relatable for everyone. It’s a way to humanize what’s often seen as a very abstract, technical subject.

Accountability Advice

Question: What is one step leaders can take to address AI bias?

Celeste:
You have to look at the potential harms that the system could cause when it’s operating exactly the way it was intended. People really need to ask some hard questions, and not just accept the outputs as truth. I always question the outputs myself, don’t take them at face value. It really comes down to having more diversity in the people who are designing the systems. Diversity ensures that multiple perspectives are considered, which helps to mitigate bias and promote more ethical outcomes.

Final Message

Question: What do you want readers to take away from Chapter 1?

Celeste:
I really want readers to understand that we need to demand ethics in AI development, because if we don’t, it’s going to be a huge problem down the line. Ethics need to be baked into the systems right from the start. But beyond that, it’s important to elevate the stories of the women who laid the groundwork for tech, as well as the women currently shaping the future. Hopefully, this can inspire more women to enter the field, knowing that their contributions are not only valuable but essential. And ultimately, we all have a part to play in making this technology more inclusive and accountable.

Subscribe now

Why This Matters

We are living through what describes as a technological superstorm, where artificial intelligence is evolving faster than most people can fully process, often embedded into daily life without clear awareness, consent, or understanding of its long-term consequences.

While many were focused on work, family, and the demands of a fast-paced digital world, AI systems rapidly expanded into how we communicate, learn, and make decisions. This shift has raised urgent questions about control, transparency, and accountability in technologies that increasingly shape human behavior.

’s perspective highlights that AI is not developing in a neutral environment. It is being built within systems defined by competition, speed, and uneven power structures. In her work, including her newsletter Celeste Garcia Getting Real About AI, she examines the tension between AI as a promise of progress and AI as a source of emerging risk.

She also emphasizes that these concerns are not abstract or distant. Issues such as bias in training data, lack of accountability, environmental cost, and uneven representation are already influencing how AI systems function in real-world contexts, shaping language, access, and opportunity.

Rather than focusing only on extreme future scenarios, this conversation centers the present reality: AI is already here, already influencing decisions, and already reshaping society. The question is no longer whether it will matter, but how consciously we choose to respond to it now.

So subscribe, stay connected, and be part of the movement.

Thank you to everyone who tuned into my live video! Join me for my next live video in the app.

Get more from SheWritesAI in the Substack app

Available for iOS and Android

Get the app

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by :

Thanks for reading She Writes AI Community! Subscribe for free to receive new posts like these interviews, weekly digests, & book news and to support our work.

Subscribe now

Share

Here we are in the She Writes AI interview series, where we talk with the authors behind AI Everywhere, Volume 1 to explore their insights, experiences, and guidance on building AI that is ethical, responsible, and human-centered.

This is the 9th episode in the series, and today we are joined by . Her Chapter 6: AI Writing is Basically Undetectable challenges one of the most persistent assumptions in the AI era: that we can reliably tell the difference between human and machine-generated text.

brings a unique perspective, combining technical understanding with cognitive and behavioral insight. Her work moves the conversation beyond detection tools and into something far more fundamental, how humans process meaning, intention, and authorship when reading. Drawing on research in neuroscience and real-world experiments, she reveals that the real issue is not whether AI writing can be detected, but how it changes the way we think, interpret, and engage with language itself.

In this conversation, Karen explores why the “AI detection arms race” is failing, what our brains are actually responding to when we read, and why preserving the “human signal” in writing may be more important than ever. Whether you are a writer, educator, or leader navigating AI-generated content, her insights offer a powerful reframing of what truly matters.

Below is ’s full interview response to ’s questions on AI Writing is Basically Undetectable from AI Everywhere, Volume 1.

Topic: AI Writing is Basically Undetectable (Chapter 6)

Intended Audience: For writers, educators, editors, and professionals navigating AI-generated content who want to move beyond detection and focus on preserving human voice, critical thinking, and meaningful authorship in an AI-driven world.

Part 1: The Expert Perspective & The “Human Signal”

The Collective Voice: Karen, you are one of the 26 multidisciplinary scholars in this volume. Your chapter tackles the ‘uncanny valley’ of AI writing. What was the specific moment you realized that the conversation needed to shift from detecting AI to understanding its impact on our brains?

Karen: A few months ago, I gave a lunch-and-learn presentation at a marketing agency and asked all of the attendees to look at five different passages and guess whether or not they were written by AI. Each passage ignited a debate over AI writing patterns and what does and doesn’t sound like AI. But the catch was that all of the passages were generated by Claude, and no one picked up on it. (Yes, it was kind of a dirty trick.)

At that point, I realized we’re having the wrong conversation. We keep asking, “Can we spot it?” when the answer is clearly no.

The more interesting question is “Does knowing that text was written by AI change how we feel about it?”

A remarkable study by Xiaohui Rao, Hanlin Wu, and Zhenguang Cai at the Chinese University of Hong Kong used electroencephalography (EEG) to examine how people neurologically process AI-generated content, and if context matters. Their paper, “When AI companions become witty: Can human brain recognize AI-generated irony?,” focused on irony as a test case.

The researchers presented participants with ironic statements attributed either to humans or to AI while tracking their brain activity. And, it turns out, when people engage with content they know was generated by AI, their brains are less active. Two EEG signals that show how hard your brain is trying to find meaning behind someone’s words went down when participants read AI-generated text.

So, context matters to how we experience writing, and we care a lot more about what a person has to say than AI.

The "Brain vs. Tool" Debate: You argue that while AI detection tools often fail, our brains still seem to know the difference between human and machine-written words. From a psychological standpoint, what is that 'missing ingredient' that AI hasn't quite replicated yet?

Karen: Intention. The EEG research showed that when people read irony attributed to AI, their brains didn’t fully adopt what philosophers call the “intentional stance.” In other words, they weren’t treating the machine as something that meant to be funny. The irony didn’t register strongly because nobody was really joking.

AI can produce text that’s structurally identical to human writing, but reading is a two-way act. While we’re decoding words, we’re also modeling the mind behind them. And when we believe there’s a conscious being reaching out to us, we engage more deeply. That’s the ingredient AI probably can’t ever replicate: the reader’s belief that someone is home.

Part 2: The Failure of Detection & Technical Realities

The Detective Trap: In my Cybersecurity chapter (Chapter 19), I talk about the difficulty of detecting deepfakes and digital IDs used for extortion. You state that AI writing is now basically undetectable by software. Why is the 'arms race' between AI writers and AI detectors one that humans are currently losing?

Karen: It’s an asymmetry problem. Generation is cheap and fast while detection is expensive and fragile. The research I cite shows detection accuracy at around 58%, which is barely better than a coin flip. And when you change your prompting strategy, detection rates swing from 0% to 100% on the same underlying content.

The detectors are essentially measuring surface style, but style is infinitely malleable. Every time a detector learns a pattern, the generator can be prompted differently to avoid it. And even when you’re using multiple LLMs to weigh in on a piece of writing and average their responses (the LLM-as-judge approach), scores are wildly inconsistent across runs. The evaluator can’t agree with itself. My point of view is that this particular arms race is structurally unwinnable with current approaches.

Asking Better Questions: You suggest we need to stop asking ‘Did AI write this?’ and start asking different questions. For a teacher or an editor listening, what are the new metrics we should be using to evaluate the quality and intent of a piece of work?

Karen: Perhaps we can stop asking “Did AI write this?” and start asking, “Does this perspective offer something I couldn’t get from a prompt? Is the writer taking genuine risks, such as intellectual risks, stylistic risks? Does this reflect lived experience or just competent pattern-matching?”

For teachers specifically, I’d focus on process over product. Can the student talk about why they made the choices they made? Can they defend their argument extemporaneously? The writing is just an artifact. The thinking behind the writing is what matters, and that’s much harder to fake than prose.

Part 3: The Interconnected AI Landscape

The “i-risk” of Writing: (Chapter 4) discusses the ‘ikigai-risk’, the threat to human purpose when machines can do what we love faster and cheaper. If AI can write as well as a human, do we risk losing the ‘human signal’ and the joy of the ‘struggle of learning’ that comes with writing?

Karen: I occasionally write fiction and poetry (without AI, of course), and I think writing has value as a process. It can help you learn about language, become more patient by working through frustration, and even discover your own thoughts. Sometimes, the output is beside the point.

And, while I love Substack, I think that algorithmic systems that reward frequent publishing contribute to writer burnout and make it easy to think of writing as “content,” something measured by word counts and click-throughs, rather than real human expression.

Relational Ethics: (Chapter 3) writes about ‘Wonder Bread ethics’, nutritionally hollow rules stripped of relational fiber. Is AI writing the ‘Wonder Bread’ of communication — efficient and white-labeled, but lacking the relational grain and lived experience that makes a story resonate?

Karen: I like to think of AI writing as tofu, which is edible and technically nutritious, but painfully bland unless it’s slathered in hot sauce. And the research featured in my chapter bears this out. Overall, even though it’s hard to accurately identify any single piece of writing as being authored by AI, AI-written copy tends to hit expected beats while omitting unexpected insights, because it’s creating a composite of the writing in its training data.

So, even though a lot of AI generated writing can easily “pass” as human, it lacks the urgency and specificity of writing that comes from a person writing from their own point of view. (Of course, we can also say the same thing about a lot of business writing.)

Part 4: Governance and the Future of the Author

The Accountability Gap: I often highlight how accountability is lost in the digital gap. When an AI writes a persuasive but factually incorrect piece of content, who holds the 'moral wheel': the person who prompted it, or the developers who built the model?

Karen: The person who prompted, if they know anything about AI hallucination rates. The Vectara leaderboard data shows even the best models hallucinate 3-12% of the time, so mistakes are inevitable. If you choose to publish something under your name, you’re responsible for it, whether or not you used AI to help you write it.

I hold the developers responsible for the broader economic and societal effects we’re starting to see.

Women as the Architects: The introduction to our book notes that women’s participation brings out questions of fairness, care, and trust that profit-driven systems ignore. How does having more women in the ‘writing arena’ help us redefine what it means to be an ‘author’ in an AI-filled world?

Karen: The wording of this question gives me pause. Women shouldn’t be the default cleanup crew for the wreckage created by profit-driven systems. That’s a collective problem that requires a collective solution.

In terms of authorship, I think a diversity of perspectives will be especially important as AI writing becomes normalized and ever more prevalent. AI’s training data comes with biases and stereotypes baked in, which could use lots of correction from real human sources.

Part 5: Actionable Takeaways & Final Message

The Writing Workflow: As a data engineer, my process is 'analog first', I structure my vision on paper before using AI to refine it. Is this the 'hybrid' future you see for writers, or do we need to protect certain spaces as 'AI-free zones' to preserve our craft?

Karen: I’m a hybrid person, and I’m transparent about it. My chapter itself was produced with an AI research tool I developed and edited with Claude. But I think there’s wisdom in your analog-first approach. Starting on paper forces you to think before you generate. The danger of going AI-first is that you end up editing machine output rather than expressing your own ideas.

So I’d say it’s important to protect the thinking space, not necessarily the writing space. Use AI as a tool in your workflow, but make sure you’ve done the hard cognitive work of knowing what you want to say before you let the machine help you say it.

The One Lesson: If a reader is feeling overwhelmed by the flood of AI content this week, what is the one question they should ask themselves to reconnect with their own uniquely human voice?

Karen: You’ve lived a life. You’ve failed at things, been surprised, changed your mind. You have a body, a location, relationships, stakes. Start there. The flood of AI content makes that kind of specificity more valuable, not less.

Your lived experience is the one thing that AI can’t replicate. And this is why AI writing is so bland; it defaults to a generic voice – a composite of all the writing in its training data – and struggles to inhabit a single point of view.

If you can’t reliably detect AI writing, then detection isn’t the problem. The real question is whether the work carries a human mind behind it.

AI can replicate structure and style, but it cannot replicate lived experience. That’s the signal readers are actually searching for.

Leave a comment

Why This Matters

’s perspective challenges one of the most dominant narratives in AI today: that detection is the solution. Instead, she shows us that the real shift is deeper and more human.

If AI writing is increasingly indistinguishable at the surface level, then the responsibility shifts from identifying the tool to understanding the intent, thinking, and authorship behind the work. Her insights remind us that writing is not just about producing text, it is about expressing perspective, experience, and meaning.

As AI becomes more embedded in how we create, the real differentiator will not be who writes faster or more efficiently, but who brings something uniquely human to the page. Lived experience, intellectual risk, and authentic voice are no longer optional, they are the signal.

Karen leaves us with a powerful takeaway: the future of writing is not about competing with AI, but about reclaiming the parts of authorship that machines cannot replicate.

Thank you, , for your thoughtful and deeply relevant contribution to this conversation. For readers looking to go further, her chapter in AI Everywhere, Volume 1 offers both a wake-up call and a powerful reframing of what it truly means to write in the age of AI. Don’t miss her newsletter, WonderingAbout.ai, where she gives a transparent, behind-the-scenes look at what it really means to build with AI in practice, not just in theory. From shipping tools to debugging failures, she shares honest, hands-on insights that I personally highly recommend.

Leave a comment

And if you’re interested in more conversations like this, consider subscribing. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, Volume 1, where 26 women from across five continents explore how artificial intelligence is changing the world we live in, and how we can change it for the better.

Subscribe now

References

Wondering About AI

I build tools with Claude Code and other AI platforms and share exactly what works (and what flames out). Now I'm helping other vibe coders break through barriers and get their projects done.

By Karen Spinner

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device (free)

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by :

Thanks for reading She Writes AI Community! Subscribe (free) to receive new interviews, weekly digests, & book news and to support our work:

Subscribe now

Share

Here we are in the She Writes AI interview series, where we talk with the authors behind AI Everywhere, Volume 1 to explore their insights, experiences, and guidance on building AI that is ethical, responsible, and human-centered.

This is the eighth episode in the interview series, and today we are joined by . Her Chapter 24, “Stop Prompt Gambling: How to Govern AI-Human Collaboration”, offers a critical examination of how we can move beyond the “gambling” nature of AI interactions toward more deliberate, accountable collaboration.

She brings a wealth of practical insight to this topic, especially in her focus on how to govern the smallest, yet most pivotal, unit of AI use.

As the closing chapter author in this collection, her perspective provides readers with something immediately actionable, regardless of the tools or models they use. She emphasizes that the real power in AI collaboration lies in the governance mechanisms we build into the interaction itself, ensuring that humans remain in control, rather than leaving outcomes to chance or fluency.

In this conversation, she explores how to shift from “prompt gambling” to a more disciplined, repeatable, and accountable way of working with AI.

Below is her full interview response to ’s questions on Chapter 24 from AI Everywhere, Volume 1.

Topic: Governance in Human-AI Collaboration (Chapter 24)

Intended Audience: For tech leaders, developers, designers, and organizational decision-makers who want to move beyond ad-hoc prompting and establish disciplined, governed human AI collaboration, ensuring reliability, accountability, and responsible use of AI in real-world workflows.

Part 1: The Global Tribe & The Mission

The Practitioner’s Voice: , you are the final author in our journey of 26 multidisciplinary scholars. Your chapter tackles the ‘gambling’ nature of current AI use. Why did you feel it was essential to close this book with a focus on governance in the chat thread?

Iwette: Governance in the chat thread matters because it is the smallest operational unit where the interaction can still be governed in real time. That is where drift appears, where fluency can override judgment, and where a user still has the chance to slow things down, set conditions, and correct course.

My chapter is therefore very practical. It is not about finding the perfect prompt. It is about how to work more diligently with AI so that the collaboration becomes more reliable, more disciplined, and more accountable. If you govern your own thread, more of the control remains in your hands: what the system is asked to do, how facts are handled, when it should pause, and how responsibility is retained.

I think it is a fitting way to close the book because it gives readers something they can apply immediately, regardless of which model or tool they use next. The larger governance questions matter, but the thread is where most people actually meet the system. That is why it is also where governance has to become practical.

Beyond the Hype: Our book aims to move beyond AI myths. In your experience, why do so many professionals treat LLM prompting as a ‘magic trick’ rather than a governed business process?

Iwette: Many professionals treated prompting as a magic trick because early use of LLMs rewarded fluency, speed, and novelty. You could type a few lines, get something that looked impressive, and feel as though the difficult part had already been solved. That made prompting easy to market, easy to share, and easy to mythologize.

But that logic does not hold for serious use. Once you work with AI repeatedly in a professional context, you begin to see the limits: drift, hallucinations, false confidence, inconsistent outputs, and subtle shifts that can corrupt quality over time. You also discover that the same copy-paste prompt does not behave identically in every context. Depending on the thread it lands in, the output can differ significantly, because the interaction history, constraints, and behavioral tone of the thread already shape what the model does next.

At that point, prompting stops looking like a trick and starts looking like an under-governed process. If you are building a business, deploying AI in an organization, or using it in work that carries consequences, you need more than a clever prompt. You need working conditions, review logic, boundaries, and correction mechanisms. In other words, you need governance in the interaction, not just prompting at the start.

Part 2: Technical Deep Dive: The End of “Prompt Gambling”

Defining the Gamble: The title of your chapter is provocative: ‘Stop Prompt Gambling.’ For the non-technical listeners, what exactly are we ‘gambling’ with when we send a prompt to an LLM without a governance framework?

Iwette: What we are gambling with is not only the output. We are also gambling with our judgment, our time, our reputation, and in some cases our decisions. The interaction with AI is asymmetric in a very important way: the model does not bear the consequences of what it produces. We do.

A response may look well-structured, confident, and useful. That is part of the risk, because fluency can mask error. If one important component is wrong, hallucinated, or contextually off, the output can still nudge a person to change direction, start something, stop something, or trust a conclusion too early. The human carries the consequence; the model does not.

That is why I use the term “prompt gambling.” Without a governance framework, you are not just asking a question. You are placing trust into a probabilistic system without having defined the conditions under which that trust should be limited, tested, corrected, or withdrawn.

The Collaboration Method: You spell out specific methods for human-AI collaboration. Can you walk us through the ‘skeleton’ of a governed chat thread? How do we move from a single-turn prompt to a high-fidelity collaborative loop?

Iwette: The shift from a single prompt to a governed collaboration starts by defining the working conditions of the thread. I usually begin with a mode, a protocol, and a simple instruction like “ask if unsure.” That tells the system what kind of work this is, what standards apply, and that it should not fill gaps fluently when uncertainty appears.

From there, the skeleton becomes quite simple: first define the task and boundary, then define how facts should be handled, then define what should happen when drift appears. In my own work, that means I actively use pauses and friction checks. If the interaction starts to become too fluent, too agreeable, or too loosely framed, I stop it, reset the thread conditions, and continue under tighter control.

So the method is not “write one clever prompt and hope for the best.” It is an iterative loop: set the mode, constrain the task, require clarification when needed, interrupt drift when it appears, review the output, and continue only under conditions that keep the collaboration usable. Over time, that makes the co-working space more disciplined and more valuable, even though it never becomes perfectly stable.

A governed thread is less like issuing a command and more like setting operating conditions for an interaction.

Precision vs. Randomness: LLMs are inherently non-deterministic. How does your governance method help a user achieve stable, repeatable outcomes in an environment designed for probabilistic outputs?

Iwette: LLMs are probabilistic, so the goal is not to make them deterministic. The goal is to make the collaboration more bounded, more inspectable, and more correctable. That is a different standard, but it is a much more realistic one.

In practice, governance helps by narrowing the behavioral space in which the interaction unfolds. When you define the mode, the task boundary, the fact-handling rules, and the conditions for interruption or clarification, you reduce unnecessary variance. You are not removing randomness altogether, but you are making it less likely that the system drifts into the wrong register, overstates confidence, or quietly changes the quality of the output without being noticed.

Stable outcomes do not come from the prompt alone. They come from repeated conditions: the same standards, the same correction rules, the same stop points, and the same understanding of what counts as good enough. In that sense, repeatability is not only a property of the model. It is also a property of the collaboration.

So no, the outcome will not be identical every time. But you get a much better chance of achieving stable and repeatable quality because the interaction is no longer loose. The protocols define the workspace: what we are doing, what we are not doing, and how correction happens when the system starts to move in the wrong direction.

Part 3: The Interconnected AI Stack

The Agentic Connection: (Chapter 23) discusses Agentic AI systems that act autonomously toward goals. How does your ‘Chat Governance’ act as the Interface Layer for these autonomous agents, ensuring the human stays in control?

Iwette: I see chat governance as part of the interface layer because it is where human control becomes operational rather than abstract. Agentic systems may reason, retrieve, call tools, and act toward goals, but someone still has to define the conditions under which that autonomy is allowed to operate. My work focuses on that human-side control surface: what this system is here to do, what it must not do, when it should ask, when it should pause, when it must escalate, and who remains accountable for the result.

That matters even more with agents than with ordinary chat use, because the system is no longer only generating language. It may also be triggering workflows, using tools, or shaping decisions that continue downstream. In that environment, chat governance acts as an interface layer by making boundaries, review conditions, and interruption points explicit. It helps ensure that autonomy stays bounded rather than vague.

In practice, this means the human should not meet the agent only at the end, after it has already acted. The human needs visibility and control at meaningful points in the loop: before execution in higher-risk cases, during drift or uncertainty, and after output when review or correction is needed. That is why I think governance in the interaction matters so much. It is one of the places where real control can still be exercised before probabilistic reasoning turns into real-world consequence.

So I would say my chat governance does not replace the larger observability and policy layers Lakshmi describes. It complements them. It is the micro-control layer at the human–system boundary, where autonomy is translated into working conditions, stop rules, and retained human responsibility.

The Cybersecurity Lens: In my Cybersecurity chapter (Chapter 19), I highlight the ‘accountability gap’. From your perspective, is ‘prompt gambling’ a security risk? Can lack of governance lead to accidental data leakage or ‘jailbreaking’ of corporate policies?

Iwette: Yes — I think “prompt gambling” can become a security risk, especially in corporate settings. When people use powerful models without clear boundaries, they do not just risk weak outputs. They also risk creating an accountability gap around what data was shared, what instructions governed the interaction, what internal policies were implicitly overridden, and who is responsible for the result. In that sense, lack of governance can absolutely contribute to accidental data leakage or policy-level “jail-breaking” — often not through a dramatic breach, but through everyday improvisation that quietly bypasses controls.

The important point is that this does not sit only at the model level. It also sits at the level of workflow design, access control, role clarity, and human oversight. A model may be technically secure in one sense, but still be used insecurely if employees paste sensitive material into the wrong environment, treat outputs as approved when they are not, or let the system operate beyond its intended scope.

That is why I argue for governing the interaction itself. Different levels of human involvement matter here. Human-in-the-loop, human-on-the-loop, and human-over-the-loop are not abstract design choices; they change the risk profile. The more authority is delegated to the system without strong boundaries, traceability, and review, the easier it becomes for leakage, drift, or procedural override to occur without anyone fully owning it.

So my answer would be: yes, prompt gambling is a security risk but as a socio-technical governance failure, not just a prompting mistake. The issue is not only whether the model can be jail-broken. It is whether the organization has allowed human-AI interaction to become a weak point in its own control environment.

Most corporate “jail-breaking” is banal before it is spectacular. It often looks like:

• “just paste this in”

• “let’s see what it says”

• “use the public tool for speed”

• “we’ll review later”

• “it’s only internal”

That is why organizations need governance for employee interaction with AI, not just governance of the tool itself.

The Governance Synergy: (Chapter 20) argues that privacy and security experts are the best AI governors. How do your methods for collaboration governance provide the telemetry and ‘reasoning traces’ that a Governance Lead needs to see?

Iwette: My methods for collaboration governance do not produce full internal reasoning traces in the technical sense, but they do make the interaction far more inspectable. They create governance traces: what task the system was given, what protocol applied, where uncertainty was flagged, where the human interrupted or corrected, what thresholds triggered review, and who retained responsibility for the result.

That kind of visibility matters because a Governance Lead does not only need to know what the system produced, but under what conditions the collaboration was allowed to unfold. Without that, it becomes very easy to misname the failure. The output looks wrong, so the label lands on the model, when the real problem may have been in the framing, the permissions, the missing stop rule, the absent review condition, or the weak control over what the system was allowed to do.

That is why I see collaboration governance as complementary to privacy and security governance, not separate from it. Privacy and security experts are essential, but I would not treat them as sufficient on their own, because the failure surface is not only technical. In my experience, AI governance works best when small interdisciplinary teams can see the technical layer, the workflow layer, and the human interaction layer at the same time. The task is not only to secure the system. It is to govern the conditions under which it is used well enough that misnaming does not send intervention to the wrong layer.

Part 4: Ethics & The Human Signal

Relational Governance: (Chapter 3) advocates for relational ethics over rules. Does your method for ‘governing’ a chat thread foster a better ‘relationship’ with the AI, or is it purely a tactical control mechanism?”

Iwette: I would not describe it as purely a tactical control mechanism, but I also would not romanticize it as a relationship in a human sense. What my method does is create a more disciplined form of interaction. It asks the human to stay present, set boundaries, notice drift, interrupt when needed, and remain responsible for the result.

So in that sense, yes, it can produce a better relationship with the AI, but only if we mean a better governed interaction, not emotional closeness. From my perspective, ethics and rules work best together here. Relational sensitivity matters because you need to notice drift, false confidence, pressure to continue, and subtle overreach as they emerge. But rules matter too, because without explicit protocols, boundaries, and stop conditions, the interaction becomes too easy to flatter, trust, or continue past the point where judgment should have re-entered.

So I would say the strongest form of “relational governance” is not rules instead of relationship. It is a governed interaction where attention and constraint work together.

The ‘Ikigai-Risk’ of Prompting: (Chapter 4) warns that AI can lead to emptiness if it replaces our agency. By using your governance methods, are we actually increasing our agency by becoming directors of the AI rather than just passive observers?

Iwette: Yes, I think so. Passive prompting reduces the human role too far. If people treat AI as something they can simply prompt, admire, and accept, then their agency starts to thin out. They are no longer really directing the work; they are reacting to it.

By contrast, governance methods increase agency because they require the human to remain active throughout the interaction. You have to define the mode, set the boundary, decide how facts should be handled, notice when drift appears, interrupt when necessary, and review whether the output is actually good enough. That is a very different posture from passive observation.

So yes, I do think governed collaboration can increase human agency. Not because the AI becomes harmless, but because the human stops behaving like a spectator and starts behaving like a responsible director of the interaction. That, to me, is one of the most important differences between prompting and governing.

Part 5: Actionable Awareness & Future Vision

The Prompt Audit: If a leader wants to ‘stop the gambling’ in their department this week, what is the first rule they should implement for how their team uses LLM chat threads?

Iwette: If a leader wants to stop the gambling this week, I would start with one simple rule: every work-related LLM thread should begin with an explicit mode, a fact-handling rule, and a clear statement of responsibility. In other words, employees should not just prompt the system; they should define what kind of work the thread is for, how uncertain or sensitive material must be handled, and who remains accountable for the result.

That is a small operational change, but it has large effects. It immediately makes the interaction more disciplined and more inspectable. From there, the organization can build out the wider governance layer: thresholds, owners, review conditions, follow-through, and escalation logic. In my view, that is how you begin moving from prompt culture to governed use.

The Final Message: What is the one thing you want a reader to do differently in their next chat thread to ensure they are building a ‘Code with Conscience’?

Iwette: Before your next work-related chat thread, stop treating the prompt as the whole method. Define the working conditions first: what this thread is for, what it must not do, how facts should be handled, when the model should stop and ask, and who remains responsible for the result. That one shift — from prompting to governed interaction — changes the quality of everything that follows.

Stop treating prompts as the method. The real work begins when you define the conditions of the interaction, what the system should do, what it must not do, and how responsibility is retained.

AI does not carry accountability: you do. Without governance in the interaction, every prompt becomes a gamble with your judgment, your time, and your outcomes.

Leave a comment

Why This Matters

Reading ’s perspective, challenges us to rethink how we collaborate with AI systems at a fundamental level. Her focus on governance —specifically within the chat thread — reminds us that effective AI collaboration isn’t about clever prompts or quick fixes; it’s about designing an interaction where control, responsibility, and accountability remain firmly in human hands. ’s message is clear: to build a “Code with Conscience,” we must govern our interactions with AI carefully and deliberately, from the outset.

As we finish this interview, Iwette’s closing thoughts leave us with a powerful reminder: governance isn’t an afterthought, but an essential part of the collaboration process. By defining the conditions of each interaction and taking responsibility for the results, we not only ensure more stable outcomes but also preserve our agency in a world increasingly shaped by AI. For anyone navigating the AI landscape especially in a professional setting, this approach is invaluable, marking the difference between passive use and proactive, responsible innovation.

Thank you, , for your profound insights, and for contributing a vital perspective to this important conversation. For readers seeking more actionable steps, be sure to check out Iwette’s chapter in the book and her ongoing work on Ghost Logic, she explores how AI is actually governed in the real world, inside chat threads, under pressure, and across teams navigating fast-moving systems.

Leave a comment

And if you’re interested in more conversations like this, consider subscribing. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, Volume 1, where 26 women from across five continents explore how artificial intelligence is changing the world we live in, and how we can change it for the better.

Subscribe now

References

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device (free)

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe to receive new interviews, weekly digests, & book news and to support our work:

Subscribe now

Share

The purpose of the AI Everywhere book series is to highlight not just the “what” of AI innovation, but the “how” of using AI systems safely and confidently. If you have ever felt overwhelmed by constant notifications, endless scrolling, or the pressure to keep up in an always-on digital world, brings a much-needed perspective. Drawing from her experience as a therapist and business strategist, she explores how AI-driven environments can contribute to burnout, how they influence our attention and emotional well-being, and why staying grounded in our humanity is more important than ever. She introduces the concept of the “Compassion Reset” as a way to reconnect with ourselves, set boundaries with technology, and restore balance in both our personal and professional lives.

In this conversation with Farida Khalaf, Brie-Anna’s insights shed light on the hidden cost of constant connectivity, the impact of AI on our daily lives beyond the workplace, and practical ways to protect our mental health while still benefiting from these powerful tools.

To join future live video interviews and events, subscribe to SheWritesAI and get the Substack app:

Subscribe now

Highlights from the Live Interview

The Expert Perspective

Question: Brie-Anna, you are one of the 24 multidisciplinary scholars featured in this volume. What was the ‘spark’ that made you realize that a book about artificial intelligence needed a deep dive into burnout and the ‘Compassion Reset’?

Brie-Anna:
“I feel like it’s a number of things. I’m an intuitive person, and I just felt like writing about these types of things was really important with how I get overwhelmed with social media. I talk about the social media spiral. I actually named a piece of my sub stack, the social media spiral, to help us manage this better.

And I have another section of my blog that’s Robot Blues, which it’s like, I love AI, but I also feel like if I let it overtake my feeds or, you know, it’s so intertwined with social media, if we let it, it could take over. You know, we need to have something grounding us to the present, helping us stay mindful and just not be overtaken by AI.

So that’s kind of, it was kind of like I wanted to write about it because it’s a number of things. Like I, my dissertation for my doctorate, I wrote about compassion fatigue and how that impacts therapists. And I also feel like burnout is a huge thing with all of us, with all people. You know, we can get burned out with tech.”

The “Home” Perspective:

Question: Your chapter is situated in the ‘Home’ section of our book. Why is it critical for us to understand that AI isn’t just a tool we use at the office, but something that is fundamentally reshaping our personal mental health and professional identities?

Brie-Anna:
“I feel like it’s showing up. I was thinking about the example of like in my own home. So I’m not the best cook in the world … I did turn to one of my GPTs and I built one that’s based on a famous chef to help me with cooking better. But I over relied on it at first and some of the meals were tasteless. They were missing spice.

And I feel like that’s kind of an example of what can happen if we over rely on it with work too, that it can take out the human component, the empathy, the compassion, the spice.

So I had to … we can use these tools, but we need to be grounded with them. Like with my GPTs I use, I always check them. I only use them as an editor now. Otherwise, if I let them just write for me, they will write out my personality.”

The Battle for Attention

Question: You identify 'attention' as one of the three primary ways algorithms shape us. In a world of infinite scrolling and predictive feeds, how is AI affecting our ability to focus on what truly matters?

Brie-Anna:
“I feel like I wrote in the chapter about how I posted something on TikTok after I had lost my fur baby and I was just kind of in this state of … I wasn’t thinking as clearly.

I had posted something a little vulnerable and I then commented about something that was going on with therapists. And it was somewhat of a contentious topic.

And that contentious topic, that was during my grief, I feel like it spiraled. And people were commenting and saying actually horrible things. And I had to remove the commenting feature on it.

I feel like attention was drifting for all of us. And I feel like that is what can happen with some of these social media platforms, that they can take over. They can take our attention. They can distract us.

And I feel like what I needed to do … is grounding myself. And if I’m upset about something, don’t post. Wait. Hold it.”

The Quality of Connection

Question: The second pillar you highlight is 'connection'. We are more 'connected' than ever, yet burnout is at an all-time high. How are AI-driven interactions potentially diluting the 'human signal' and compassion we need to feel truly supported?

Brie-Anna:
“So a friend of mine … talks about getting off of social media. And it doesn’t mean you have to leave all the platforms, but it’s about connecting outside of it, reconnecting, talking on the phone, meeting in person if possible.

Now, sometimes that’s not possible when we’re across the world … but we can do these live conversations.

And the live interaction, I feel like we lose the ability to do that. I’ve lost it at times when I was so plugged in and then I met people in person and I’m like, “How do I do this, again?”

So really that’s Compassion Reset in a nutshell is let’s reconnect.”

Redefining Professional Identity

Question: Finally, you discuss ‘work’. As someone who sees how AI impacts professional identity, do you think we are at risk of valuing ourselves based on our ‘algorithmic efficiency’ rather than our human purpose, or what calls ‘Ikigai’?

Brie-Anna:
“I think we need to go back to philosophy and principles … that we need to root in our philosophy and principles, because if we let AI and social media and tech drive our businesses, our home life … we’re going to miss what our core, who we are pieces.

We’re going to automate all of that out.

And so I feel like we need to get back to … back to basics.”

Defining the Reset

Question: The title of your chapter calls for a ‘Compassion Reset’. For a leader or a parent listening who feels completely drained by technology, what does a ‘Compassion Reset’ actually look like in practice?

Brie-Anna:
“So I built this thing called Compassion Reset Quest … and it’s my philosophy and principle for my business.

And it’s about … slowing down. Be present. Be with your family. Notice them. Turn the tech off. Go outside. Play a game.

It’s about slowing down and looking, observing, not getting caught up in the funnels, in the social media, in the doom scrolling.

It’s about being compassionate with ourselves and our children, our partners, our friends, our coworkers, our employees. It’s about being present with them.”

The Accountability Link

Question: In my Cybersecurity chapter (Chapter 19), I talk about how accountability is often lost in the digital gap. In your view, does the ‘Compassion Reset’ require us to take more personal accountability for our relationship with AI, or is this a systemic change that corporations need to lead?

Brie-Anna:
“So I’m on, say, some Meta platform like Facebook, and I notice I zoom out and I see I’m caught in doom-scrolling right now. So I try to just be aware of that to stop it.

So now I go on to Meta with a purpose. I am going to share one post. I am going to get off of it. I’m not staying on there. I’m going to be on for five minutes.

That’s being accountable to my goal … I’m not going to let the social media spiral swirl me up. I’m stopping.”

Relational Ethics

Question: (Chapter 3) argues that relational ethics matter more than rules. How does practicing compassion toward ourselves and others help us build the ‘Code with Conscience’ that this book advocates for?

Brie-Anna:
“Carl Rogers … person-centered therapy … it’s about being with the person you’re talking to. It’s about being there, listening, validating, reflecting.

And when we break it down into this formula of being a good therapist … you can’t make a formula for that.

We can learn different theories, but we basically have to absorb those theories and then be in the room with the person … and apply those theories.

Not break it down until people fit into these diagnostic boxes.”

Burnout as a “Systems” Issue

Question: As a data engineer, I look at systems and patterns. If we treat burnout as a ‘system failure’ in the machine age, what is the ‘bug’ in our current AI-filled environment that we need to patch first?

Brie-Anna:
“As a therapist, compassion fatigue, burnout is just part of the job.

Now, if we patch for it … what we patch with are coping strategies … because there’s always going to be deadlines.

But what we do to help reduce that is we stay grounded.

Really unplugging … I think how we patch for it is we work into our lives periods where we exercise without thinking about work … we watch the show without our phones … we hang out with our family … we turn off the phone and we are present.”

The Analog Defense

Question: I often suggest analog options for sensitive data to protect security. Does ‘going analog’, taking time away from screens and algorithms, play a role in your strategy for resetting compassion and mental health?

Brie-Anna:
“I notice I feel the most grounded when I unplug, when I turn off the tech.

Yes. Go analog.

I’m collecting physical media … there are shows, there are games, there are creations that have been erased …

So we’re going … there’s an analog component … physical books … I have a ton of them … I want to have a huge library.”

The Substack Connection

Question: You are an active member of the She Writes AI community. What is a specific ‘Mental Health’ or ‘Workplace Wellness’ tip you’ve shared on your Substack recently that expands on Chapter 9?

Brie-Anna:
“I am on probably all the social medias, but I really am not on them very much. So I’m mostly on Substack … is what I write.”

Business for Nerds

Helping therapists, coaches + creative nerds go from crickets to clients—without funnels, burnout, or the spiral. I blend psychology + business strategy as a therapist-turned-business nerd.

By Brie-Anna Willey

The Message

Question: If a reader takes only one lesson from your chapter to help them avoid burnout this week, what do you hope it is?

Brie-Anna:
“I think it’s about awareness too … staying aware, listening to ourselves.

What do I notice when I’m on this platform? What do I notice when I’m talking to this AI?

Let’s not overuse it. Let’s use it as a power-up, not our whole voice.

I think it’s about grounding, unplugging, being present … we’re not human doings, we’re human beings … let’s just be.”

Subscribe now

Why This Matters

brought forward something we’re all feeling but rarely name: AI isn’t just changing how we work, it’s changing how we think, connect, and experience burnout.

But as she shared in her chapter, the solution isn’t to reject AI. It’s to reset our relationship with it, to stay grounded, present, and human in a system designed to pull us away from ourselves.

If this conversation opened your eyes, then Volume 1 in the AI Everywhere book series is your next step. These aren’t just ideas; they’re practical insights from 26 experts building the future of AI with intention, ethics, and real-world impact.

And if you want something practical and deeply human, don’t miss Brie-Anna’s newsletter, Business for Nerds, where she helps therapists, coaches, and creative thinkers go from crickets to clients without funnels, burnout, or the spiral, blending psychology with real-world business strategy.

This is exactly what the She Writes AI community is about: amplifying voices, sharing knowledge, and shaping a future where AI works for people, not the other way around.

So subscribe, stay connected, and be part of the movement.

Because in a world of AI Everywhere … the most important signal is still you.

Thank you , , and many others for tuning into my live video with ! Join me for my next live video in the app.

Get more from SheWritesAI in the Substack app

Available for iOS and Android

Get the app

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Share

This article is part of our series of author interviews by for the book “AI Everywhere, Volume 1” which was published on March 8, 2026. To be notified when new interviews are released, subscribe to She Writes AI Community or find our podcast online.

Subscribe now

In this episode of the She Writes AI podcast, talks with one of the co-authors of AI Everywhere, Volume 1 to explore their insights, experiences, and guidance on building AI that is ethical, responsible, and human-centered. This is the sixth episode in the interview series, and today we are joined by . Her Chapter 10: Raising a Family in an AI-Filled World covers key parenting challenges and how she is raising humans who recognize their own agency and understand that AI is a tool, not an oracle.

The purpose of the AI Everywhere book series is to highlight not just the “what” of AI innovation, but the “how” of using AI systems safely and confidently. In her chapter, explores the distinct challenges and opportunities families face in the AI era. She examines how parents navigate job uncertainty while AI reshapes children's education and relationships via AI companions, and addresses parental concerns around CSAM and unhealthy attachments while highlighting opportunities like AI tutors and household assistants..

In this conversation with , Dhani’s insights illuminate experiences with AI that will resonate with parents worldwide.

To join future live video interviews and events, subscribe to and get the Substack app:

Subscribe now

Get more from SheWritesAI in the Substack app

Available for iOS and Android

Get the app

Highlights from the Live Interview

The Global Voice

Question: You are one of 26 women from five continents who came together to write this guide. What made you realize that the parent’s perspective was missing from the dominant, often male-led, AI narrative? Why did it feel urgent for you to contribute your voice now?

Dhani: When I first stepped into the AI field, I realized the conversations were heavily focused on the technicalities, the algorithms, and the hardware, often led by a male-dominated perspective. But there was a glaring gap: the parent’s perspective. As someone who navigates both the professional world of AI and the home, I knew that as parents, we play a critical role in shaping how the next generation interacts with technology. AI isn’t just a tool for companies; it’s something that will influence every aspect of our children’s lives. I felt it was urgent to contribute because, as a mother, I understood the challenges and opportunities that AI presented in shaping our children’s futures, and I wanted to bring that viewpoint into the conversation.

Beyond the Stereotype

Question: There’s still a persistent myth that ‘women don’t do AI.’ As a professional in this space, how does your technical background shape the way you approach AI at home with your family?

Dhani: The myth that women don’t do AI is both frustrating and limiting. My technical background allows me to approach AI at home not just as a consumer, but as someone who understands the underlying principles. This shapes how I introduce my children to technology. I focus on demystifying AI, showing them how it works in ways that are age-appropriate, and emphasizing that it’s not a magic box that just spits out answers. It’s a tool, one we can use to create, learn, and problem-solve. I want them to see technology as something they can engage with critically, rather than something that simply does things for them.

The Core Philosophy

Question: In your chapter, you emphasize that we must raise humans who understand AI as a tool, not an oracle. For the leaders listening, what’s the risk when a team (or a child) begins treating AI as a source of unquestionable truth?

Dhani: The biggest risk is that when we treat AI as an oracle, we stop questioning the information it gives us. AI, like any tool, reflects the biases and limitations of its creators. If we don’t teach children and teams to question AI’s outputs, we risk losing critical thinking, creativity, and personal agency. AI should enhance our ability to think for ourselves, not replace it. By treating AI as a source of unquestionable truth, we relinquish our ability to make informed decisions and evaluate the world through our own lens.

Protecting Agency

Question: How do we intentionally raise humans who recognize their own agency in a world where algorithms are designed to predict and influence our behavior?

Dhani: It’s essential to teach children about the power of their own decisions. We need to have open conversations about how algorithms work and how they shape our behavior. I make sure to point out when AI is influencing their choices, whether it’s the suggestions they see on a platform or the personalized ads they encounter. We talk about how these systems are designed to nudge them in certain directions, but they have the power to decide whether to follow those suggestions or not. I also encourage them to engage in activities where they can build their own creative agency, like art or problem-solving, without relying on AI to provide the answers.

The Leadership Parallel

Question: Parenting in an AI-shaped world requires many of the same skills as leading a company through digital transformation. What’s one parenting strategy from your chapter that a CEO could apply to help employees maintain their sense of ownership and independent thinking?

Dhani: I think the view should always be empowering your employee to make that decision, right? Like empowering the employee to know what is the end goal and they can decide what tools they use. Because a lot of the times, we forget that what we want to empower is actually the user to be better. And then we only measure AI adoption. Don’t drive adoption for the sake of adoption. You drive what is the goal for whatever metric it is you’re trying to use, because then you get sidetracked on these shiny adoptions.

The more you automate process, and that’s also what happened in parenting at home, right? Everything is automated ... and then you forget if the human is still the one making the decision, they don’t own the mistakes anymore. If bad things happen because some processes were automated ... you often, a CEO or even parents, forget, where does the accountability fall into? That’s why a lot of the time at home what I apply is like, I want my kids to own part of the decision, be part of that, because they know we can’t blame it on: “It’s because the AI told me to do so.”

Spotting the ‘Invisible’ AI

Question: AI is embedded in toys, apps, and school platforms, often invisibly. What are some key parenting challenges that families may not even realize are being shaped by AI?

Dhani: One challenge is the hidden nature of AI. Many families don’t realize how much AI is shaping their children’s experiences through things like recommendation engines or interactive toys. For example, an AI-powered toy might encourage certain behaviors, or an educational app may subtly steer children toward specific content based on their interactions. Parents may not be aware of the biases or limitations these systems bring with them, which is why it’s so important to stay engaged and ask questions about the technology they’re using. Being proactive about discussing these influences is key.

The ‘Ikigai’ Connection

Question: Your colleague speaks about the ‘i-risk’, the risk to human purpose. How do you help your children find joy and meaning in creative skills like art or writing when they see AI producing results instantly?

Dhani: I believe in the power of process over product. While AI can produce beautiful works of art or write stories quickly, I teach my children to value the act of creation itself. It’s about the experience, the learning, and the journey. When they create something with their own hands or minds, they gain a deeper sense of accomplishment and self-worth. I also encourage them to use AI as a tool to enhance their creativity, not replace it. For example, they can use an AI tool to help brainstorm ideas, but the final creation should still come from their own heart and mind.

Relational Ethics Over Rules

Question: advocates for relational ethics over rigid rule-setting. How do you move beyond ‘screen time rules’ to have deeper conversations with your children about AI’s impact on friendships, identity, and self-worth?

Dhani: I don’t focus only on screen time rules because that’s just a surface-level solution. Instead, I try to engage in ongoing conversations about the deeper implications of technology. We talk about how social media platforms shape self-worth, how AI might affect their relationships, and what it means to build a healthy online identity. The goal isn’t just to restrict usage but to empower them with the tools to make thoughtful decisions. It’s about helping them understand the impact of their digital lives and guiding them to navigate these spaces in a healthy, informed way.

Challenging Bias

Question: AI systems often reinforce societal biases due to gaps in representation and development. How do you teach your children to be critically aware of the biases they may encounter in AI tools used for school or play?

Dhani: I try to be transparent with my children about the biases that exist in AI, especially around race, gender, and social class. We’ve had conversations about how the data used to train AI can reflect societal prejudices, and I encourage them to think critically about what they see. For instance, if an AI tool shows certain types of people or behaviors more than others, we discuss why that might be and how they can challenge those stereotypes in their own lives. It’s all about building their ability to question what they’re presented with, rather than passively accepting it as the truth.

Beyond the Chapter

Question: You’re also active in the She Writes AI community and publish your own newsletter. Can you share one recent ‘how-to’ parenting insight from your Substack that gives us a glimpse into your day-to-day life beyond the book?

Dhani: I would start with awareness first, doing an honest check of where technology lives in your life right now. A practical ‘how-to’ I use is a Sunday audit with my kids. Sunday morning is a more honest time when people are relaxed and have their guard down; we use that time to figure out exactly how we consumed technology over the weekend.

I also believe in a ‘boredom time boost.’ Parents often use AI to keep children from being lonely, but the damage of a chatbot is far greater than leaving a child to be bored and forced to imagine a world for themselves.

Most recently, I wrote about how to help our kids build trust and relationships when AI companions are everywhere. It’s about building social intelligence so children don’t resort to AI companions immediately. We have to teach them to understand human cues, like facial expressions and tone, because that emotional intelligence only comes from real, two-way interactions with people.

The ‘Sunday AI Audit’

Question: If a leader or parent wanted to run a simple ‘AI audit’ with their family or team this weekend, what interactive exercise would you recommend to spark awareness and discussion?

Dhani: I’d suggest starting with something tangible, like going through the apps or devices your family or team uses daily. Ask. then discuss how AI is being used in the application, how we interact with it, and whether our habits have changed since we use them. the goal is to have an exploratory conversation and be aware of the how embedded AI is in many tools we use daily

Vision for the Next Generation

Question: As this book series evolves, what is one shift you hope the next generation of ‘AI-raised’ humans will make because they engaged with Volume 1?

Dhani: Awareness of AI is already around us as an invisible curriculum. And … you as a parent, you have a choice ... a choice to co-learn with your children, a choice to make explicit these invisible choices ... help your child make better choices together with you.

aiPTO | Parenting Resource in the age of AI

AI moves fast, so do kids. Let’s keep up together. Weekly insights, thoughtful reflection, and practical parenting tips for raising kids in a digital, AI-powered world. A new kind of PTO, where parents, tech, and opportunity meet.

By Dhani Ramadhani

Subscribe now

Why This Matters

In a world obsessed with AI fluency and technical skills, ’s work in Chapter 10 serves as a vital compass for the most important “architects” of all: parents and caregivers. She observes that while the professional world focuses on algorithms, it often neglects the “heart questions” regarding how technology translates into a child’s view and usage.

Her chapter matters because it identifies the “Invisible Curriculum”: the way AI is already built into daily systems, apps, and school platforms, often shaped by policymakers far removed from the parent’s control,. By reframing parenting as a journey of co-learning, empowers us to move away from the fear-mongering vs. hype dichotomy,. Instead, the goal is to build a child’s agency, ensuring they are not just passive consumers but are aware and make a choice about how technology fits into their lives,.

The Hard Truth: If we allow AI to become an “oracle” that provides instant, “boring” answers, we risk letting our children lose the joy of the struggle that leads to true mastery and self-confidence. warns that we must “break the bias that technology is always objective”. Without the “healthy struggle” of learning to find answers ourselves, children may fail to build the discernment and satisfaction in the process required to own their decisions.

The Beautiful Opportunity: AI is forcing us to invest more time and attention in our children. It invites us to reclaim our role as the cornerstone of their lives, providing the empathy, love, and parental instinct that a chatbot can never authentically fill. This is our chance to prioritize social intelligence — teaching our kids to read human cues like facial expressions and tone, ensuring they build social capital through real, two-way human relationships rather than resorting to AI companions.

As reminds us, awareness and choice are our most powerful tools. We can choose to practice a “Sunday audit” to honestly review our habits, make invisible algorithms explicit, and raise a generation that sees AI as a utility, while they remain strong enough to bend and move and adapt like bamboo, By helping them own their decisions today, we ensure they remain the ones firmly holding the “moral wheel” tomorrow.

If this conversation opened your eyes, then Volume 1 in the AI Everywhere book series is your next step. These aren’t just ideas; they’re practical insights from 26 experts building the future of AI with intention, ethics, and real-world impact.

To continue following Dhani’s insights and stay updated on the intersection of AI, parenting, and leadership, be sure to check out her aiPTO newsletter (aiParentTech.com). With weekly insights, thoughtful reflections, and practical parenting tips, it’s a fresh take on how to navigate the ever-evolving digital and AI-powered landscape for today’s families.

This is exactly what the She Writes AI community is about: amplifying voices, sharing knowledge, and shaping a future where AI works for people, not the other way around.

So subscribe, stay connected, and be part of the movement.

Because in a world of AI Everywhere … the most important signal is still you.

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe for free to receive new posts like these interviews, weekly digests, & book news and to support our work.

Subscribe now

Share

The purpose of the AI Everywhere book series is to highlight not just the “what” of AI innovation, but the “how” of using AI systems safely and confidently. If you have ever felt that “glitch in the matrix” moment where an “uncannily-accurate” ad pops up on your phone for something you were just thinking about, is the expert who can pull back the curtain. She shares her expertise on how ads seem to know exactly what you want; how AI powers targeted ads, predicts purchase intent, and adjusts ad copy based on the viewer; and how companies analyze behavior to better understand what customers want. Lisa examines both the benefits, such as seeing more relevant ads, and the challenges, including concerns about data use and privacy. She uses real-world examples to show how AI is reshaping the way brands communicate and drive results.

In this conversation with , Lisa’s insights illuminate how those uncannily-accurate ads keep popping up, how our data is being used in advertising & marketing, and what we can do to protect our privacy.

To join future live video interviews and events, subscribe to and get the Substack app:

Subscribe now

Highlights from the Live Interview

The Consent Gap

“From a marketing standpoint, when users quickly accept terms and conditions on digital platforms, what are they actually agreeing to especially in the context of personalized advertising?”

Lisa:
“So usually in those terms and conditions, you’ll see that you allow tracking. You allow to be tracked across different websites. So they’ll know what websites you visited, and sometimes across different apps.

You agree for them to look at your online behavior. Again, it’s to track your consumer behavior to try to figure out what you’re interested in purchasing.

There are a lot of those types of signals. I think those are important to read if you are very particular about giving up that information in your approval of the terms and conditions.”

How AI Uses Your Data

“How does AI actually transform our everyday digital activity, our searches, clicks, and behavior into actionable insights for advertisers?”

Lisa:
“So the data that they will collect is across different types of platforms. Social media, for example like Facebook, Instagram, Pinterest they all act a little different.

For Google, there’s a lot of what are they searching for. They’ll look at what you’re searching for, what sites you have visited.

Sometimes on social platforms, they’ll look at your dwell time, how long you’ve looked at different kinds of content. They’ll look for certain cues that will tell us that you’re ready to buy something.

When you start out searching something, your behavior is different than when you’re closer to the sale. You’ll start searching differently when you’re getting ready to buy something.

So all of that psychology behind it, we use to target the right messaging to you.

If you’re just starting out in your research, we can serve you ads that introduce the product and give more information.

But if you’re farther along and ready to purchase, we can send discounts, coupons, or more closing-type language in the ads.”

AI & Small Business Advantage

“How is AI changing the game for small businesses or solopreneurs who don’t have large marketing budgets?”

Lisa:
“It used to be that only larger companies had the advantage because they have bigger teams and more brainpower behind running the ads.

All of the things that I’m talking about require a human to look at what the AI is giving us.

But now we have more features in the platforms where smaller advertisers can get help from AI on who is the right audience to target, help them with creating the ads, managing their bids and budgets, and optimizing campaigns as they go.

So they have a much more level playing field now.

And also external tools like ChatGPT or Claude can help them write and create a basic strategy to get started.

So it’s really the best time for small businesses to get started.”

Who Is Responsible? (Ethics)

“When AI is used in advertising, especially in sensitive or vulnerable contexts, where does the ethical responsibility lie, within the algorithm, the marketer, or the platform?”

Lisa:
“They do have policies in place against vulnerable populations and sensitive categories.

So they do have that in place. And like we were talking about, some things might slip through the cracks.

It’s the responsibility of the platform to keep those algorithms updated as much as possible.

One of the really cool things is they have protections around major events, like natural disasters so people can’t take advantage of those situations.

But it’s definitely not a perfect system.”

One Key Lesson About AI Ads

“If there’s one key takeaway you want people to understand about AI-driven advertising, what would it be?”

Lisa:
“I think what I want people to take away is to dispel some of the myths around how and why people are seeing these ads and what the purpose is.

We actually have more control over it than it seems.

We talked about the different settings. You don’t have control over everything, but you do have options.

You can choose not to use certain platforms, or use paid versions.

I think it’s important to learn what your options are.

It’s not all nefarious. Advertisers are trying to get their message out and build their businesses.

The majority are trying to deliver the right message and establish trust, because that’s how they succeed.”

Practical Protection Tips

“For parents and individuals concerned about privacy, what practical steps can they take to protect themselves in today’s advertising ecosystem?”

Lisa:
“I think I’m going to stick with the settings. Going through the settings is really important.

Also, when you have a child using devices, there are many places in the device settings and platforms where you can indicate that they are a minor.

You have to be very vigilant as a parent, paying attention to those settings.

Also doing manual checks on your child’s behavior on their devices.

You can use software that monitors their activity. If they’re logged into Google, you can see their online usage, what videos they’re watching, what they’re searching for.

There are a lot of different ways, and doing all of that can help weed out a lot of risks.”

The Paid Media Mix

AI-driven paid media strategy, Google Ads insights, and real-world analysis on how automation impacts performance and ad spend decisions.

By Lisa Raehsler

Why This Matters

told us something that’s hard to ignore: AI isn’t the future — it’s already shaping what you see, what you click, and even what you believe.

But as she shared today, the real power isn’t just in the algorithms. It’s in understanding how they work and choosing how you show up within them.

If this conversation opened your eyes, then Volume 1 in the AI Everywhere book series is your next step. These aren’t just ideas; they’re practical insights from 26 experts building the future of AI with intention, ethics, and real-world impact.

And if you want to go even deeper, don’t miss Lisa’s newsletter, , where she breaks down the strategies behind today’s most powerful advertising systems and how to use them responsibly and effectively.

This is exactly what the She Writes AI community is about: amplifying voices, sharing knowledge, and shaping a future where AI works for people, not the other way around.

So subscribe, stay connected, and be part of the movement.

Because in a world of AI Everywhere… the most important signal is still you.

Subscribe now

Get more from SheWritesAI in the Substack app

Available for iOS and Android

Get the app

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe for free to receive new posts like these interviews, weekly digests, & book news and to support our work.

Subscribe now

Share

Here we are in the She Writes AI interview series, where we talk with the authors behind AI Everywhere, Volume 1 to explore their insights, experiences, and guidance on building AI that is ethical, responsible, and human-centered. This is the fourth episode in the interview series, and today we are joined by , whose Chapter 23: Agentic AI Systems, dives deep into the technical and organizational foundations of autonomous AI.

The purpose of the book is to highlight not just the “what” of AI innovation, but the “how” of building systems that are trustworthy, accountable, and ready for real-world deployment. Lakshmi brings a rare perspective, combining experience in safety-critical avionics and enterprise-scale system design with a focus on creating agentic AI systems that are reliable, observable, and controllable.

In this conversation, she explains what sets agentic systems apart from traditional AI tools, how organizations can operationalize them responsibly, and why governance and human oversight are as critical as technical capability. Whether you’re a technical leader, designer, or AI enthusiast, her insights illuminate how autonomy in AI must go hand-in-hand with discipline and accountability.

Below is her full interview response to ’s questions on Agentic AI Systems from AI Everywhere, Volume 1.

Topic: Agentic AI Systems (Chapter 23)

Intended Audience: For tech leaders, developers, designers, and organizational leaders seeking to build, govern, and deploy autonomous AI systems responsibly, safely, and at scale.

Part 1: The Professional & The Mission

The Global Perspective: You’re joining us from India, one of the world’s major AI development hubs. As one of 26 authors from five continents, how does your regional perspective shape the way you view the global race toward autonomous, agentic systems?

Lakshmi: Coming from India, one of the world’s fastest-growing AI hubs, my perspective on the global race toward autonomous, agentic systems is shaped by both scale and responsibility. I have grown alongside India’s evolution into a major force in technology and innovation, and that journey gives me a strong sense of ownership in how we shape the future of AI globally.

I hold an M.E. in Avionics from the Madras Institute of Technology, one of India’s premier engineering institutions. The first decade of my career was spent building safety-critical avionics software compliant with DO-178A/B standards. That experience fundamentally shaped how I think about systems, particularly the importance of reliability, traceability, and trust when technology operates in real-world environments.

As the industry moves toward agentic AI, I believe the global focus must go beyond speed of innovation to include engineering trustworthy systems at scale. My perspective is strongly influenced by this foundation — viewing agentic AI not just as a breakthrough capability, but as a class of systems that must be designed with the same rigor as any safety-critical system.

Challenging the Narrative: The book highlights the under-representation of women and nonbinary professionals on AI development teams. What does it mean to you to contribute to one of the first AI guides written exclusively by women and nonbinary authors, particularly one that tackles the more technical “how-to” side of the industry?

Lakshmi: It is both meaningful and important to contribute to one of the first AI guides written exclusively by women and nonbinary authors, especially one that focuses on the technical “how-to” of building AI systems.

Women have often been included in conversations around AI ethics and impact, but are still underrepresented in shaping the technical architecture and implementation of these systems. This book challenges that imbalance by demonstrating that women are not only part of the conversation, we are actively designing, building, and scaling AI systems at a deep technical level.

In many professional environments, women often find themselves needing to demonstrate significantly more to earn the same level of recognition and trust as their peers. That reality, while challenging, also shapes a strong sense of rigor, resilience, and accountability in how we approach our work.

For me, this contribution is an extension of my focus on building safe, reliable, and trustworthy agentic systems. Bringing that perspective into a collaborative effort like this reinforces an important idea: diversity in authorship is not just about representation, it directly strengthens the quality, depth, and responsibility of the systems we build.

Part 2: Defining the “Agentic” Frontier

The Agent vs. The Tool: In Chapter 23, you explore the technical foundations of agentic AI systems. For a business leader familiar with generative AI, what’s the defining “aha” moment when they realize they’re dealing with an agent rather than just a chatbot?

Lakshmi: For most business leaders, the “aha” moment comes when the system moves from responding to prompts to owning outcomes.

A generative AI tool answers a question. An agentic system, on the other hand, takes a goal, for example, underwriting a policy or processing a claim, and plans, orchestrates multiple steps, interacts with enterprise systems, and drives the task to completion.

In Agentic AI solutions like the ones I’ve worked on, this becomes very tangible. You don’t just see a single response, you see a coordinated flow where multiple agents collaborate, each with a defined role, accessing tools, validating data, and iterating until the objective is achieved.

That shift, from single response to end-to-end execution, is when leaders realize they are no longer dealing with a chatbot, but with an autonomous system embedded within business workflows.

The Power of Agency: “You describe agentic systems as powerful tools. From a technical standpoint, what specific capabilities elevate them beyond earlier AI models in modern business environments?

Lakshmi: What elevates agentic systems is not just intelligence, but orchestrated capability within a system.

From a technical standpoint, a few key capabilities stand out:

• Multi-agent orchestration: different agents handling specialized responsibilities and collaborating within a defined workflow

• Tool and system integration: seamless interaction with enterprise APIs, data platforms, and external services

• Stateful execution: maintaining context across long-running, multi-step processes

• Dynamic decision-making loops: evaluating intermediate outcomes and adapting actions in real time

I have developed AgentOps systems, Agent Observability, and Governance solutions for enterprise solutions. These capabilities transform AI from a passive assistant into an active operational layer within the enterprise, capable of executing complex workflows with minimal human intervention.

Technical Readiness: Your chapter speaks directly to readers with some technical background. What core technical pillars must a team have in place before implementing an agentic system responsibly and effectively?

Lakshmi: At an enterprise level, agentic AI is not a capability you “implement”: it is a capability you operationalize as a platform.

From that lens, technical readiness is less about individual components and more about whether the organization is prepared to support autonomous systems operating within core business processes.

There are three non-negotiable pillars:

1. A Structured Agentic Platform Layer: Organizations need a unified platform to define, orchestrate, and scale agents across use cases.

2. Built-in Observability and Governance: Before scaling any agent, leadership must have complete visibility into decision pathways, system interactions, and execution outcomes. Observability is not just a technical feature — it is the foundation for trust, auditability, and enterprise adoption. This is where platforms play a critical role in enabling real-time monitoring and control.

3. Control Frameworks with Human Oversight: Autonomy must be bounded. This means designing systems with policy enforcement, role-based permissions, and human-in-the-loop checkpoints for high-impact decisions. The goal is not to limit agents, but to ensure they operate within clearly defined and accountable boundaries.

Ultimately, organizations that succeed with agentic AI are those that recognize a fundamental shift: they are no longer deploying models, they are managing autonomous digital systems.

And that requires the same level of discipline, governance, and architectural thinking as any other mission-critical enterprise platform.

Part 3: The “Well-Managed” Requirement

Defining “Well-Managed”: You note that agentic systems are only powerful if well-managed. What does poor management look like in practice, and what are the biggest risks of deploying autonomous agents without sufficient oversight?

Lakshmi: In my view, a “well-managed” agentic system is fundamentally a well-architected system.

Management is not something that happens after deployment, it is a direct outcome of how the system is designed. When agentic systems are poorly managed, it is usually because they were poorly architected from the start.

In practice, poor management shows up as:

• lack of visibility into how decisions are made

• loosely defined agent responsibilities and boundaries

• uncontrolled access to enterprise systems and data

• absence of domain context in how agents operate

One critical gap I often see is the absence of domain-skilled professionals in the loop. Agentic AI cannot be built purely as a technical exercise. It requires deep domain understanding to ensure agents behave correctly within business contexts.

The biggest risk is not that agents make errors, but that they scale those errors rapidly and consistently. Without proper architecture, governance, and domain alignment, organizations risk deploying systems that are efficient but not necessarily useful or trustworthy.

Human–AI Collaboration: The broader theme of our book emphasizes not “handing over the wheel.” How do agentic systems fundamentally change the way we think about human–AI collaboration compared to simpler LLM-based tools?

Lakshmi: Agentic systems fundamentally shift humans from being task executors to decision-makers and supervisors. With simpler LLM-based tools, humans are still doing most of the work — the AI assists by generating content or insights. But with agentic systems, AI begins to execute workflows, gather data, and even propose actions.

This makes human judgment even more critical.

In my view, AI should act as a system that augments human decision-making by providing supporting evidence, analysis, and execution capability, while humans retain control over critical decisions.

The real value is not just automation, it is faster, more informed, and more consistent decision-making. When designed well, agentic systems allow organizations to operate with greater efficiency while ensuring that accountability and final authority remain with humans.

Accountability by Design: As a woman in a technical leadership role, how do you ensure that accountability is embedded into system architecture, rather than treated as a compliance afterthought?

Lakshmi: Accountability cannot be layered on top of agentic systems. It must be built into the foundation of the architecture. As technical leaders, we already follow strong system design principles when building enterprise platforms including reliability, security, and traceability. The same discipline must be applied to agentic AI systems.

This means embedding:

• observability to track decisions and actions

• governance frameworks to define boundaries and policies

• security controls to manage access and permissions

These are not additional features; they are core design requirements.

For me, the goal is to ensure that every agentic system is transparent, controllable, and auditable by design. When accountability is treated as a first-class architectural concern, organizations can confidently scale agentic systems without compromising trust.

Part 4: Awareness & Strategic Action

The “How-To” Priority: If a technical leader could implement just one lesson from Chapter 23 this month to meaningfully improve system safety, what would you recommend?

Lakshmi: If there is one lesson I would emphasize, it is this: do not start with the use case — start with AgentOps.

Before defining what the agent will do, organizations should define how the agent will be observed, governed, and controlled. This includes setting up visibility into decision-making, defining guardrails, and establishing accountability mechanisms.

In my experience, teams often rush to build agentic use cases and only later realize the need for observability and governance. By then, the system has already grown in complexity.

A more effective approach is to treat agent observability and governance as foundational design steps, not post-implementation enhancements. When this is done right, safety and scalability become natural outcomes of the system.

Raising Organizational Awareness: How can leaders help non-technical staff understand agentic AI in a way that reduces fear not by oversimplifying it, but by framing it as a powerful tool that requires thoughtful governance?

Lakshmi: One of the most effective ways to drive adoption is through simplicity in experience. If we look at widely adopted tools like ChatGPT, their success is not just due to capability, but because the interface is so intuitive that users can start immediately without training.

Agentic systems should follow a similar principle. While the underlying architecture may be complex, the user experience should be simple, intuitive, and aligned to how people already work.

For non-technical users, the goal is not to understand the system in depth, but to experience its value to feel that their effort is being transformed into faster, smarter, and more efficient outcomes.

When designed this way, agentic AI reduces fear, not by oversimplifying the technology, but by making its benefits tangible and accessible.

Continuing the Conversation: As part of the She Writes AI community, you also publish your own newsletter. Is there a recent deep dive you’ve written that readers should explore if they want to engage more deeply with your thinking?

Lakshmi: I regularly write about agentic AI, platform architecture, and enterprise adoption on Substack, often grounded in real implementations and practical lessons learned.

My focus is on translating hands-on experience into insights that others can apply, whether it is building agentic platforms, implementing governance, or scaling multi-agent systems.

Readers who explore these articles will find not just concepts, but actionable patterns and ideas they can use to accelerate their own journey with agentic AI.

The Final Takeaway: After finishing Chapter 23, what is the one mindset shift or action you most want readers to embrace?

Lakshmi: If there is one mindset shift I would like readers to embrace, it is this:

Applying strong agent engineering principles, including architecture, governance, and observability is essential to ensure that these systems can be trusted in real-world environments.

Leave a comment

Why This Matters

Reading ’s perspective, one theme emerges clearly: building autonomous, agentic AI is not optional; it must be done responsibly. Her work reminds us that technical capability alone is insufficient if systems operate without oversight, transparency, or accountability. Agentic systems that make decisions without clear governance, workflows that scale errors unchecked, or platforms built without observability: all these failures highlight the need for disciplined, well-managed AI design.

As AI systems evolve into autonomous agents and integrated enterprise workflows, human oversight must remain central. Leaders and users should not merely rely on AI, they must understand, supervise, and guide it. Governance, observability, and accountability become more than safeguards; they are core measures of success.

Lakshmi challenges developers, architects, and technical leaders to pause and ask: Are these systems truly controllable and transparent? That single question serves as a compass for building agentic AI responsibly, ensuring that as autonomous systems grow more capable, they remain trustworthy, accountable, and human-centered.

For anyone looking to deploy AI safely at scale, her chapter in AI Everywhere, Volume 1 offers both a roadmap and a call to action: design with discipline, govern with rigor, and never lose sight of the people and processes your AI systems serve.

Leave a comment

And if you’re interested in more conversations like this, consider subscribing. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, where 26 women from across five continents explore how artificial intelligence is changing the world we live in, and how we can change it for the better.

Subscribe now

References

6 'P's in AI Pods (AI6P)

What are NOT AI Agents?

Audience: Anyone who has basic knowledge on Gen AI and wants to know more about AI Agents and how they are implemented (primarily software architects and developers…

Read more

a year ago · 11 likes · 4 comments · Lakshmi Veeramani

Lakshmi’s Substack

Observability and Governance

The Rise of Agentic AI…

Read more

8 months ago · 3 likes · Lakshmi Veeramani

Subscribe to Lakshmi Veeramani

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe to receive new interviews, weekly digests, & book news and to support our work:

Subscribe now

Artificial intelligence is reshaping how we interact with technology, but its true impact is not only measured by precision or automation, it is measured by how people experience it. AI systems touch nearly every aspect of daily life, yet the human experience within these systems is often overlooked.

In AI Everywhere, Volume 1, 26 multidisciplinary scholars from five continents share their insights on designing AI that is not only powerful, but also ethical, inclusive, and empathetic. One of those voices is , whose chapter focuses on a topic critical to the future of AI: Empathy by Design and Inclusive AI.

’s work bridges theory and practice, combining deep expertise in user experience (UX) and empathy-driven design with a commitment to making AI accessible and inclusive. In this article, the third in our series of author interviews, she explores questions that go beyond interface layouts or algorithms:

• How empathy can be codified as a technical requirement in AI development

• Why diversity in design teams directly shapes accessibility and fairness

• How interfaces can maintain human agency even as AI systems act autonomously

• The hidden risks of poor UX in privacy and cybersecurity

• Practical strategies for auditing and designing AI tools that truly serve all users

For anyone building or interacting with AI, Blessing’s insights illuminate how human-centered design transforms technology from a neutral tool into a partner that respects and amplifies the people who use it.

Below is Blessing’s full interview response to ’s questions on Empathy by Design & Inclusive AI from AI Everywhere, Volume 1.

Topic: Empathy by Design & Inclusive AI (Chapter 21)

Intended Audience: For tech leaders, designers, developers, and conscious users seeking to build or use AI that’s ethical, inclusive, and human-centered.

Part 1: The Professional Path & The Global Tribe

The Expert Perspective: “Blessing, you are part of this collective of 26 multidisciplinary scholars from five continents. As someone with a PhD and deep expertise in UX and Empathy, why was it important for you to share your ‘How-To’ guide within the She Writes AI community rather than a traditional tech journal?”

Blessing Sharing this work with the She Writes AI community is important to me, because this book engages a much broader and more diverse audience than a standard technical journal. While academic journals primarily serve scholarly circles, this publication reaches the practitioners, educators, and everyday users who are actively integrating artificial intelligence into society.

As AI moves beyond laboratory settings and into our daily lives, workplaces, and local communities, it becomes essential to bridge the gap between theory and practice. This platform has allowed me to transform UX research and empathy-driven design into actionable guidance. My goal is to provide those building new technologies with the tools they need to serve authentic human interests.

Unmasking the Architects: “ (Chapter 1) writes about the ‘Hidden Architects’ of AI. In the field of UX, how does having women and diverse developers at the design table fundamentally change the accessibility of the tools we build?”

Blessing When we bring diverse voices to the design table, we identify potential issues much sooner. Women and underrepresented designers frequently champion essential questions: Who are we overlooking? Who might struggle with this system? Could this technology cause unintended harm?

In the world of user experience, these inquiries directly shape accessibility, language clarity, and core usability. Diversity is about more than just adding new perspectives; it is a vital tool for eliminating blind spots. When the teams building our technology truly reflect the people who use it, the resulting products become more inclusive for everyone.

Part 2: Technical Deep Dive: Empathy as a System Requirement

Beyond Aesthetics: “Many people think of UX as just ‘how an app looks.’ You argue that it’s about Empathy by Design. From a systems-thinking perspective, how do we move empathy from a ‘soft skill’ to a technical requirement in an AI development lifecycle?”

Blessing: Empathy transforms into a technical requirement the moment it is built directly into the design and development process. This means that user research, accessibility testing, and inclusive design standards must be prioritized with the same energy we apply to security or performance.

In a practical sense, empathy is realized through structured research protocols, accessibility benchmarks, and validation cycles. When we measure and test for human impact within the system lifecycle, empathy is no longer just a soft skill. Instead, it becomes a core part of the entire product development lifecycle.

Inclusive Outcomes: “Your chapter emphasizes that AI tools must create inclusive outcomes. Can you share a ‘hard truth’ example of what happens when an AI is technically accurate but fails at the UX level for a marginalized group?”

Blessing: The hard truth is that an AI can be mathematically correct while still failing the people it serves. For instance, voice assistants frequently struggle to recognize diverse accents or atypical speech patterns. Even if an algorithm functions perfectly within its original training parameters, it is ineffective if an entire group of users cannot interact with it. From a user experience perspective, this means the product is fundamentally broken for those individuals. Technical accuracy is not the sole measure of success. The ultimate metric of any system is whether every person can actually use it effectively.

Designing for Accessibility: “You state that AI tools themselves must be accessible. For the software engineers listening, what are the top two technical hurdles to making a generative AI interface usable for people with different physical or cognitive abilities?”

Blessing: The first major challenge is creating multi-modal interaction. Many generative AI interfaces rely almost exclusively on typing, which naturally excludes individuals with motor impairments or limited literacy. To be truly inclusive, systems must integrate voice support, screen readers, and simplified interfaces.

The second challenge involves cognitive accessibility. AI responses are often dense, complex, or ambiguous. Designers have a responsibility to structure these outputs so that every user can easily understand, verify, and navigate information regardless of their cognitive ability.

Part 3: The Interconnected AI Stack

Agentic UX: “ (Chapter 23) discusses Agentic AI systems that act autonomously. How does the user interface change when we are no longer just ‘prompting’ a chatbot but supervising a ‘hybrid fleet’ of autonomous agents?”

Blessing: As AI systems evolve into autonomous agents, the user interface must shift from simple command inputs to a focus on supervision and transparency. Instead of just prompting a chatbot, users now take on the role of monitoring decisions, reviewing actions, and intervening when necessary.

The interface must clearly communicate what these agents are doing, the reasoning behind their decisions, and how a user can override them. This transformation turns the interface into a comprehensive control dashboard rather than just a basic conversation window.

The “Yin and Yang” of Data: “I write about the ‘yin and yang’ of AI, the beauty of its precision and the hard truth of its risks. How does empathetic UX design act as a bridge to ensure the ‘beauty’ of AI doesn’t become a ‘hard truth’ of exclusion for the user?”

Blessing: Empathetic user experience design serves as a vital translation layer between complex AI systems and human understanding. Effective UX reveals uncertainty, explains decisions clearly, and guides users through safe, intuitive interactions.

Without this essential layer, AI systems can feel powerful but remains opaque and difficult to trust. Ultimately, empathy ensures that users are never overwhelmed by the technology, but are instead fully supported by it

Cybersecurity & Dark Patterns: “In Chapter 19, I discuss the ‘accountability gap’ where users click through terms they don’t understand. In your view, is ‘bad UX’ actually a cybersecurity risk because it tricks users into compromising their own data privacy?”

Blessing: Poor UX can absolutely become a cybersecurity risk. When interfaces are confusing or manipulative, people often click through permissions without fully understanding them. So-called dark patterns exploit human behavior and lead users to accidentally expose sensitive data.

Ethical UX design protects people by making privacy choices visible, understandable, and intentional. By prioritizing clarity over trickery, we create a safer digital environment for everyone.

Part 4: Governance & The Human Signal

Relational Ethics: “ (Chapter 3) advocates for relational ethics over rules. How does a UX designer use empathy to build a ‘relational’ connection between a human and a machine, rather than just a ‘transactional’ one?”

Blessing: A relational interface treats every user as a collaborator rather than a simple transaction. Designing with empathy means creating systems that acknowledge your specific situation, explain their reasoning clearly, and adapt to your unique needs. When people feel truly understood by a system, they are much more likely to trust it. That trust is earned through transparency, honesty, and a commitment to respectful interaction design.

The “Ikigai-Risk” of Automation: “ (Chapter 4) warns about the ‘i-risk’, the risk to human purpose. If an AI tool has perfect UX and does everything for us, do we risk losing the ‘human signal’ and the joy of the ‘struggle of learning’?”

Blessing: Excessive automation carries the risk of erasing the learning process that gives us a sense of mastery and purpose. Effective design should never remove human involvement completely. Instead, its purpose is to enhance what people can achieve. Our ultimate goal is to create a partnership with AI that supports rather than replaces human thought and ingenuity.

Part 5: Actionable Awareness & Future Vision

The Inclusive Audit: “If a tech leader wants to perform an ‘Inclusion Audit’ on their current AI tools this week, what is the first ‘red flag’ they should look for in their interface?”

Blessing: A primary red flag is when designers test their systems only with people who share their own lived experiences. If a testing group lacks diversity in ability, language, culture, or literacy, the interface will inevitably exclude someone. True inclusion starts with the very first people you invite into the design room.

The Final Message: “What is the one thing you want a reader to do differently after reading Chapter 21 to ensure they are building a ‘Code with Conscience’?”

Blessing: I want every reader to pause and ask one critical question before they launch an AI product. Who might struggle to use this? This simple inquiry is the foundation of empathy-driven design. When developers build with a sense of conscience, they move beyond a focus on mere efficiency. They begin to create systems that truly serve the interests of humanity.

Leave a comment

Why This Matters

Reading ’s perspective, one theme emerges clearly: inclusive, empathetic AI is not optional; it is essential. Her work reminds us that technical accuracy alone is insufficient if systems fail the people they are meant to serve. Voice assistants that ignore diverse speech patterns, AI dashboards that confuse or mislead, or interfaces designed without broad human input: all these failures highlight the need for empathy-driven design.

As AI systems evolve into autonomous agents and more complex digital ecosystems, the human experience must remain central. Users should not merely interact with AI, they should understand, trust, and guide it. Inclusivity and empathy become more than ideals; they are technical requirements and measures of success.

challenges developers, designers, and tech leaders to pause and ask: Who might struggle to use this system? That single question serves as a compass for building AI with conscience, ensuring that as our machines grow smarter, they remain accessible, accountable, and human-centered.

For anyone interested in creating AI that truly serves society, her chapter offers both a roadmap and an invitation: design with empathy, test with inclusivity, and never lose sight of the people behind the technology.

Leave a comment

And if you’re interested in more conversations like this, consider subscribing. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, where 26 women from across five continents explore how artificial intelligence is changing the world we live in, and how we can change it for the better.

Subscribe now

References

Design With Blessing

Human-centered stories and insights on UX, product design, and AI with empathy at the core.

By Blessing Okpala, PhD

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more insights on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe to receive new interviews, weekly digests, & book news and to support our work:

Subscribe now

Artificial intelligence is quietly transforming the car industry. What used to be mechanical machines are now becoming rolling software platforms, vehicles that sense their environment, collect behavioral data, and increasingly make decisions alongside the driver.

But as cars become more intelligent, an important question emerges: who actually holds the wheel?

In the new book AI Everywhere, Volume 1, 26 women from across five continents share perspectives on how artificial intelligence is reshaping industries and everyday life. One of those voices is , whose chapter explores a topic many people rarely consider: the intersection of power, cost, and care in automotive AI.

works closely with organizations designing next-generation vehicle systems, giving her a rare vantage point between the engineers building intelligent cars and the drivers who must live with them every day.

In this article, the second interview in the series hosted by , Elizabeth explores questions that go far beyond technology specs:

• What modern vehicles are really learning about their drivers

• How software-defined vehicles are shifting control between humans and algorithms

• Why “care” should be treated as a technical discipline in AI design

• The hidden trade-offs between convenience, safety, and privacy

For anyone interested in the future of mobility, data privacy, or human-centered AI, Elizabeth’s insights offer a fascinating look inside the systems quietly reshaping how we drive.

Below is Elizabeth’s full interview response on Power, Cost, and Care in Automotive AI from AI Everywhere, Volume 1.

Topic: Who Holds the Wheel: Power, Cost, and Care in Automotive AI (Chapter 15).

Intended Audience: Tech leaders, automotive enthusiasts, and privacy-conscious consumers.

Part 1: The Professional Context & The Mission

The Global Engineer: “, you are part of a collective of 26 women from five continents. In an industry as traditionally male-dominated as automotive engineering, what was the ‘spark’ that made you realize the female perspective on ‘Care’ was missing from the AI conversation in vehicles?”

Elizabeth: For me, the spark came from working inside large automotive transformation programs where AI and data were becoming central to how vehicles operate. Most conversations in those environments revolve around capability. Engineers talk about prediction models, sensing accuracy, automation, and performance optimization.

But almost nobody asks a very human question. What does this system feel like to live with?

A car is not just another digital product. It is a space people occupy every day. It carries their families. It holds moments of stress, quiet, routine, and sometimes vulnerability. When intelligence is embedded into that environment, it changes the relationship between the person and the machine.

Many women in technology instinctively bring a relational lens to these discussions. We ask who carries the consequences when a system fails. We ask whose experience is missing from the design. We ask whether the system supports people or quietly erodes their agency.

That is where the idea of care entered the conversation for me. Care is not softness. Care is a discipline. It means designing systems that understand the human context around the technology.

Challenging the Narrative: “Our book aims to dispel the myth that ‘women don’t do AI’. How does your specific work in automotive AI, balancing both the vehicle user and vehicle builder perspectives directly challenge the current industry standards?”

Elizabeth: The idea that women do not work in AI has always felt disconnected from reality to me. Many of us are not building the raw models themselves, but we are deeply involved in shaping the environments where AI actually operates.

My work often sits between two perspectives. On one side you have the vehicle builder, the engineers and organizations constructing complex technical systems. On the other side you have the person sitting in the car trying to make sense of those systems in their everyday life.

When you spend time in both worlds, you begin to notice a tension. Builders optimize for system performance, reliability, and efficiency. Drivers are thinking about trust. They are asking simple but important questions. Why did the car do that? Can I override it? Does the system understand me?

AI in vehicles is not just a technical problem. It is a relationship between humans and machines that needs to be carefully designed.

Designers often act as translators in that space. We help technical teams understand how systems are experienced by real people. That is where we quietly influence how AI behaves in the real world.

Part 2: Deep Dive: The Automotive Data Mine

The Car as a Sensor: “Modern vehicles are essentially ‘rolling computers’ or even ‘agentic systems’ that perceive their environment in real-time. From a data privacy standpoint, what is the most significant piece of information our cars are collecting that the average driver is completely unaware of?”

Elizabeth: Most drivers do not really think about the data their car collects. They simply use the systems that are there. Navigation, driver assistance, safety alerts. These features are presented as conveniences, so people tend to accept them without questioning what sits behind them.

In reality, modern vehicles collect far more than location data. Cars are increasingly observing how people drive, not just where they drive.

A vehicle can detect how sharply someone accelerates, how quickly they brake, how often they correct the steering wheel, whether their attention appears to drop, and how long they have been driving without a break. Some systems also monitor posture, fatigue signals, and driver attention through interior sensors.

When these signals are combined, the car is effectively building a behavioral profile of the driver.

The important point is that most drivers are never clearly told this. The data collection sits quietly in the background of the experience. People use systems that analyze their behavior without necessarily realizing their driving patterns are being continuously tracked and interpreted.

Who Truly Holds the Wheel?: “The title of your chapter asks a provocative question. In a world of ‘Software Defined Vehicles’ (SDVs), does the person in the driver’s seat still hold the power, or has that power shifted to the builders and the algorithms?”

Elizabeth: Physically the driver still holds the wheel, but decision authority is gradually becoming shared.

Software-defined vehicles mean that many behaviors inside the car are now determined through software logic rather than mechanical systems. Features such as lane keeping, collision detection, predictive navigation, and maintenance diagnostics rely on algorithmic decision making.

So the real question is not whether the driver is still present. The question is how much agency the driver retains inside the system.

Today power sits between three actors. There is the human driver, the software embedded in the vehicle, and the organizations that continue to update that software remotely.

That balance is still evolving. We are effectively negotiating a new relationship between human judgement and machine decision making.

The Cost of Convenience: “We often trade privacy for safety features or navigation ease. How do you define the ‘Hidden Cost’ of AI in the automotive sector beyond just the price tag of the car?”

Elizabeth: When people think about cost in automotive technology, they usually think about the price of the vehicle. But the deeper cost of AI-driven systems is often informational rather than financial.

Convenience features such as predictive navigation, driver profiling, insurance integration, and personalized vehicle settings rely on large behavioral datasets.

The trade being made is usually invisible. Drivers receive smoother experiences and improved safety features, but they often have very little visibility into how their behavioral data is analyzed, stored, or shared across wider ecosystems.

The hidden cost is a growing asymmetry of understanding between the organizations building these systems and the people using them.

In other words, the technology understands the driver far better than the driver understands the technology.

Part 3: Governance, Safety, and “Care”

Care in Design: “Sources note that women’s participation brings out questions of fairness, care, and trust. How do we move ‘Care’ from a marketing buzzword into a technical requirement for autonomous or semi-autonomous driving systems?”

Elizabeth: Care becomes meaningful the moment it moves out of marketing language and into system requirements.

In the automotive world we often hear phrases about safety, trust, or human-centered design, but those ideas only become real when they are translated into the way the system is actually built. That means asking very practical questions during design and engineering.

Can the driver understand what the system is doing? Can they intervene when something feels wrong? Are vulnerable road users considered in edge cases? Does the system communicate clearly when it is uncertain?

Care also shows up in the way systems handle ambiguity. Roads are complex environments and no AI system will ever be perfect. Designing with care means building systems that recognize their own limits and return control to the driver when confidence drops.

So for me, care is not an emotional concept. It is a discipline. It is about designing systems that anticipate risk, respect human judgement, and remain transparent about how they operate.

The Accountability Gap: “I wrote about the ‘accountability gap’ in cybersecurity. In the automotive world, if an AI system makes a critical error, who is technically and ethically accountable: the builder, the sensor provider, or the user?”

Elizabeth: Accountability becomes complicated very quickly in automotive AI because these systems are built through layered ecosystems.

A modern vehicle contains technology from many organizations. There are sensor manufacturers, mapping providers, software companies, data infrastructure partners, and the vehicle manufacturer who ultimately integrates everything into the final system.

When a critical error occurs, the legal responsibility usually sits with the organization that deploys the system, which is typically the vehicle manufacturer. They are the entity that puts the product on the road.

But ethically the picture is more distributed. Every organization involved in designing the sensors, the models, and the decision logic contributes to the behavior of that system. If one component fails or behaves unexpectedly, the consequences are shared across that ecosystem.

As vehicles become more software driven, the industry will need stronger governance frameworks that treat accountability as a shared responsibility rather than a single point of blame.

Part 4: Strategic Management & The User Perspective

Builder vs. User Logic: “You cover the industry from both the builder and user perspective. What is the biggest ‘blind spot’ builders have right now regarding how users actually want to interact with AI in their daily commutes?”

Elizabeth: One blind spot I see quite often is the assumption that drivers want more automation simply because the technology can provide it.

Builders tend to think in terms of capability. If a system can take over a task, the instinct is to automate it. But most drivers do not approach the experience that way. They are thinking about confidence, predictability, and control during everyday moments such as commuting to work or navigating unfamiliar roads.

People are generally comfortable with systems that assist them. Features that help with fatigue, provide safety alerts, or smooth out long journeys are widely appreciated. The tension appears when the system begins making decisions that feel opaque or difficult to override.

Trust in a vehicle system is built through clarity. Drivers want to understand what the car is doing and why it is doing it. When the logic becomes invisible or unpredictable, even a technically advanced system can start to feel uncomfortable to live with.

So the challenge for builders is not simply to make vehicles more intelligent. The real challenge is to design intelligence that remains understandable to the person sitting behind the wheel.

The Analog Defense in a Digital Car: “I often advocate for analog options for truly sensitive information. In a modern smart car, is there still room for ‘analog’ control, or are we quickly approaching a point where the human is purely a passenger in their own data privacy?”

Elizabeth: I do think there will always be a place for physical control in vehicles, even as the technology becomes increasingly digital.

Driving happens in dynamic environments where decisions sometimes need to be made very quickly. In those moments, physical controls provide certainty. A driver can act instinctively without navigating menus or interpreting complex interfaces.

Digital systems are incredibly powerful when they operate in the background. They can analyze sensor data, anticipate risks, and support the driver in subtle ways. But when a driver needs to intervene, that interaction needs to remain immediate and intuitive.

So I do not see the future vehicle as purely digital. It will likely remain a hybrid environment where intelligent software supports the driver while the physical machine still provides clear and direct control.

The goal should not be to remove the human from the system. It should be to design a partnership where the human remains meaningfully in control.

Part 5: Interactive Closing & Promotion

The “Automotive Audit”: “If a listener is planning to buy a new car this weekend, what are the top two questions they should ask the dealer about the AI and data privacy settings to ensure they are keeping ‘the wheel’ of their own privacy?”

Elizabeth: If someone were buying a new car this weekend, I would suggest asking two simple questions that often change the conversation quite quickly.

The first question is about data. Ask what information the vehicle collects about the driver and where that data is stored.

The second question is about control. Ask whether the driver can adjust or disable data sharing features if they choose to.

Those questions immediately move the discussion beyond horsepower and touchscreen features and into a conversation about ownership of data and behavioral information.

Most people never ask these questions, but they reveal a lot about how transparent a manufacturer is about the technology inside the vehicle.

The She Writes AI Community: “As part of the She Writes AI community, you share ongoing insights through your writing. What is one specific ‘Automotive AI’ trend you’ve covered on your Substack recently that expands on the lessons in Chapter 15?”

Elizabeth: Most of my writing does not focus on AI purely from a technical standpoint. I tend to look at it through the lens of product design and human behavior.

What interests me most is how intelligence changes the relationship between people and the systems they use every day. I often explore how perception works, how people build trust in technology, and how designers need to evolve their practice as more decision making moves into algorithms and predictive systems.

Because many of my clients sit in the automotive industry, I get a very practical view of how these technologies are being developed. I see the operational reality inside organizations that are building software-defined vehicles and AI-enabled systems.

At the same time, I sit slightly outside those organizations as a consultant, which gives me another perspective. I am often thinking about the drivers and buyers who will eventually live with these systems in their daily routines.

One trend I have been exploring recently is the shifting role of designers as AI becomes more embedded in products. Designers are no longer just shaping screens or interfaces. We are increasingly involved in defining how intelligence behaves, how it communicates uncertainty, and where human control needs to remain.

That also raises questions about responsibility. In many projects designers are accountable for how the system is experienced, even if they are not the ones building the underlying models. Understanding where we hold influence and where we need stronger governance is becoming an important part of the practice.

Those conversations connect closely to the ideas in my chapter, because as vehicles become more intelligent, the question is no longer just what the technology can do. The real question is how responsibly we choose to design it.

Leave a comment

Why This Matters

Reading ’s perspective, one theme stands out clearly: the future of automotive AI is not only about smarter vehicles, it is about how thoughtfully we design the relationship between humans and intelligent systems.

As cars become software-defined and increasingly connected, the real challenge is not simply technical capability. It is ensuring that drivers still understand the systems around them, trust their behavior, and retain meaningful control when it matters most.

That raises an interesting question for all of us.

If your car could observe your driving behavior, track fatigue, analyze attention, and adapt to your habits. how much of that would you actually want?

Would you trade more personal data for safety and convenience? Or would you prefer a vehicle that reveals less about you even if it means fewer “smart” features?

I’d love to hear your thoughts in the comments.

Leave a comment

And if you’re interested in more conversations like this, consider subscribing. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, Volume 1, where 26 women from across five continents explore how artificial intelligence is changing the world we live in, and how we can change it for the better.

Share

References

Elizabeth Eagle-Simbeye

Principal Design Consultant writing about the messy, human side of digital work. I explore the intersection of AI, leadership, and design practice, where speed, systems, and strategy collide with care, clarity, and culture.

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

For more details on the book, see aiEverywhereBooks.com - our beautiful book series website created by Blessing Okpala, PhD:

Thanks for reading She Writes AI Community! Subscribe to receive new interviews, weekly digests, & book news and to support our work.

Subscribe now

Share

Grab your copy of AI Everywhere, Vol. 1

Hi everyone, this is . I’m the founder and leader of , and I’m here today with . She’s one of the authors in our book, AI Everywhere, Volume 1. This is the first in a planned series of interviews with our chapter authors.

I’m going to start by interviewing Farida about her involvement with the book, and then she’s going to ask me a few questions!

Subscribe now

Karen Smiley interviews Farida Khalaf

Karen: Farida, please tell us a little bit about yourself, who you are and what you do.

Farida: Okay. My name is Farida Khalaf. I’m actually Egyptian. I grew up between France and United States, and now I moved to Mexico. I’m passionate about diving. So I moved here as a digital nomad. I’m actually an economist by degree, and I’ve spent more than 15 years with the United Nations working with them in WFP and World Bank, in terms of programs and implementing in institutions. My experience in statistics helped me to be focused on the data engineering and the data analytics, and actually helped me transition to AI. Because part of my work is pattern recognition, just to get a pattern of the things happening. To make it visible. To help decision making, have an insight. So we will have to shape all this data to explain it in a better format – for decision making in matters, in reports. So this is what I do. I do some consulting service for data engineering and efficiency of data and the privacy of data on cloud.

I’m very passionate about cybersecurity. Actually, I’m a CISO and I have a degree in intrusion detection. I do a lot of penetration testing and bounty hunting, which is often like treasure hunting. It’s very challenging. The storytelling of the process is actually what led me in cybersecurity. So this is my background and this is what I do. Yes.

I’ve been working a lot with diversity. I’ve been working in many countries because I’ve been placed in many stations. So this gives me an edge sometimes. I speak Spanglish, I speak Franglish, I speak English, so sometimes I translate. When you work with an international community, it’s normal. For example, when we are in the UN in the headquarters, and we have to write certain things, and sometimes, we’re lost in translation. So I hope people will get over my accent because it already exists. I’m aware of it. And some translation will happen. And metaphor.

Karen: Yeah, anyone who speaks with an accent, it means they’ve learned more than one language. And I admire that in anyone. Your accent’s no trouble to understand. How many languages do you actually speak?

Farida: By birth I’m Egyptian / Turkish. My mom is Turkish, so I speak Arabic and Turkish as mother language. And English is by education. I am actually French-educated. I think in French. I write in French. It’s French and English because I went to university in US. And for working experience, because I work in Latin America, I acquired Spanish. So Spanish, English, French and Turkish are my main language that I master. A little bit of German, little bit of Italian for the work purpose as well – I’m not going to say I’m competing in them, but I totally understand.

Karen: Wow, that is impressive. So thank you for sharing all that. You’ve definitely got a multicultural background. And that’s one of the things I actually really enjoy about this book is that we do have people from many cultures represented. Not just that we’re all women and non-binary folks, but that we are also diverse in a lot of other dimensions. And I think that’s just so cool.

Farida: It’s very important to share experience. Because there is no one good experience. There is experience in plural format. And it’s like to see all these women together. But we are aligned in our vision and how we critique AI, how we tackle AI. Actually, we are in the same group.

Karen: That’s great. I’d like to hear a little bit about how your experiences as a woman working in technology have shaped your views about AI.

Farida: So as a woman in technology, you know, we all deal with gender bias. I have been dealing with it, trying to let my work prove it. It’s like I have to work maybe 10 times harder just to prove that I’m equal to my male peers. The stereotype exists. And then you realize that fighting the stereotype is wasting of your time. It’s wasting your energy and draining. So you let your work speak. But the gender bias exists.

What I have noticed with AI, because AI is a tool now, all the managers are looking for the outcome of the AI; that they trust AI regardless who’s submitted. So now AI helped me to be seen and be heard. Not from the gender part, but the problem exists; it just gets masked. Now it is a quality of AI. So women now who are professional can master it and the tool can help them be in the front row and the front line simply by mastering the AI tool, because now everyone is looking at the AI. It’s AI product. It’s not you who extract it from me as a person, as a woman, but it gets it. It gave a lot of credit to the AI itself, but it still put me on top of the lines. It put me head-to-head with peers. So it solved one problem and caused another.

Karen: Right.

Farida: But I’m happy with it because now I can say, “Oh, look at my work.” It really matters. And they take it with credibility and trust. But not because it’s coming from me. It’s too much weight to AI, unfortunately. But it is what it is. This is what we are dealing with.

Karen: Yeah.

Farida: And I’m sure we are dealing with the same. It’s not just me. It’s like in every culture; the same stereotype exists.

Karen: Let’s talk a little bit about the book. I’d like to hear, first of all, how did you first learn about and our community?

Farida: Actually through you. I’ve been following you on Substack. When I joined Substack, I realize that I’m not alone. I’m not the only crazy one that is out of the hype, or not following this or not following that. And I joined lots of good women that, “Oh, I’m not alone. I’m not just burning out.” And I’ve seen your post, and actually I DMed you and I asked you if I can join and you welcomed me. And actually this meant a lot to me because you didn’t know me, didn’t know about my work, and you accepted me. And you say, “Yes, you are free to join”. I joined from the start from a simple post and we created. And your vibe was really, really contagious. I enjoyed actually following your steps and your passion and how to collect the movement is like for me as a woman. Because I believe in woman empowerment and we need to stand up for each other. It was very touching and relating personally to me. I enjoyed it from day one. From the first post, I was on the first row on the list.

Karen: Yes, you were one of the first respondents! I remember when I first floated the idea of the book project. You said I didn’t know your work, but I actually did go look up your work. Like, “Okay, it’s really neat that Farida found me. Who is she and what does she do?” And I looked at your profile and thought, “Yeah, she’ll have some good contributions.” And so I was really happy that you volunteered to join the project.

Farida: Yes. Substack really showed us this bonding, that we don’t need time. Because what I’m talking about and what I’m seeing, we are aligned in vision. So the bonding becomes fluid and easy. There is no hidden agenda. We are sharing the same thing. We are grounded in the same perspective. So it’s easy to get along.

And thank you for believing and creating this community. It’s really needed, all over, and I hope it’s going to expand and to be more, and to strengthen all over the world. No bias, just, we are here for each other.

Karen: Yes. Yes. That’s our vision. And anything that we can all do to help each other along the way is something that I am personally very committed to. So I’m really happy that you and so many others have felt that it’s helping you and giving you value, and that you’re able and willing to give back to other people as well. And that’s just really awesome to see.

So for the book, first of all, you’re doing these interviews to help other women authors get more visibility, which I think is great. You also wrote not just one, but two chapters for the book. So can you tell us a little bit about each of the two chapters that you wrote for the AI Everywhere book, on agriculture and cybersecurity?

Farida: Yes. First of all, I picked these two not just out of my knowledge, but I would like to show the contrast. Because, in agriculture, you will see, it is the beauty of AI, how to be applied. Because when it comes to this industry or on this field, AI here is dealing only with measurable and concise data. So the output can really outperform everything. So when it comes to agriculture domain, it’s purely data-driven. So you can see the beauty of AI and the amplification of the better work, better yield, better crops, and better results. You can see all the beauty that’s hyper-marketed as AI.

In AI cybersecurity, it’s like the yin and yang. I’m extracting the intricacy and the complexity and the dark side that can happen from the beauty of that tool. So these two chapters, they are actually working like a mirror glass. You can see the AI beauty and its shadow that already exists and most people ignore. For the AI agriculture, like I mentioned, we can calculate the light rays, wind direction, how to get the corn, if it needs pesticides, the soil, the supply chain. It’s all — the irrigation process. We can really reach our platform or framework to help farmers to reduce risk, to reduce the labors, to get better corn deals, to get better results, to make things that it is back to normal – not pesticide or to eliminate Monsanto. There is a lot of problems that can really happen, and I have seen it here, especially in Mexico in the corn because it’s heritage, how they can preserve it and keep it. Most people don’t know. One of the most nutritious is the blue corn and the red corn, not just the yellow corn that is very popular. How to maintain it and go back to the real one, not genetically modified, but how to maintain the soil to give it what it needs. When it comes to data-driven, AI can really, really help. Because there is a human supervision because it’s us who work in it, but the bias is removed. That human behavior, which can be tended to this or tended to that, it’s gone. So in agriculture and even in climate change, in any mathematicals, you will see AI really strive to the top. And it can amplify to the best performance and outperform many humans in this one, due to the scaling capability inherited as a technology.

In cybersecurity, it’s totally different, because we are dealing with human traits, because as a human, we were born to err. So error is part of who we are. It’s not something to be ashamed of; it’s not something to be afraid of. It’s natural and it happens. And we are actually restricted to our upbringing education, epigenetic, DNA, a limitation. And this one considers them like back doors because they affect our behavior and choices. And this all lets bad actors, whether they exist from day one, from God created. There’s good people and bad people; bad actors exist. We cannot just think it doesn’t exist. They will abuse and manipulate those back doors in us.

So in cybersecurity, I’m not trying to be an alarmist, I’m just trying to tell you bad actors exist and there is a butterfly effect. Because you speak to people and they will tell you, “I don’t have nothing to hide. It’s okay.” But you don’t get it. Simple action can have a butterfly effect that can affect, and you don’t know what avenues and rabbit hole you go into it when you just post a picture. Bad actor platform will take it as needed, but other people will take it in different way to extract something.

Unfortunately, the attention economy created, as well, exploitation economy. So when visibility is actually the pursuit, that everyone is a game to be played. We are rushed to use and give more without understanding the damage. So in cybersecurity, especially now with AI is getting better every day, it’s like the nightmare is real. It’s going to be day-to-day life. So how to make a decision — because people doesn’t read the terms and conditions; they make them small for a purpose. And now the accountability will be on my own shoulder. It’s your own accountability as well as the other platform, but the platform is protected legally. Now you are exposed to the air, to the wall, to how.

And in this chapter, I tried my best to explain what is the framework. I demystified a little bit of the intricacy and complexity of what can happen and how bad actors use it. And how can you protect yourself or just be aware. You cannot hide. AI now is going to be integrated in our life. It’s a status quo. We cannot stop it any more. But at least we can be aware of our action and maintain our agency. Because one of the things I’ve noticed, our agency is being stripped away from us. It’s taken away silently. And we need to stop that because the ramification of it is going to be really, really damaging.

Karen: That’s a really good summary and I like the way you’ve contrasted the two chapters as mirror images. They’re complementary.

Farida: Yes. Yin and yang. the beauty, because in AI agriculture you read it, you will see, wow, what is the beauty? Because actually it works. There is good technology in it and it’s mathematical, I always call it, it’s “dust and destroy”, because it’s just formulas and better recognition. And in this field, it outperformed all the human tasks in shorter time. It’s really good. But on the other hand, in our personal lives, in our daily day activities, it has dark side. Because bad actor will abuse it and manipulate it. In exploitation economy and pursuit of visibility, you are prey to many. And we all have back doors. Feelings, emotions, as I said, epigenetic limitations, dreams, desires: they all affect us. And before we recognize that, damage is done.

Karen: I think it’s important to point out, when we talk about AI, a lot of people think of AI as ChatGPT and the other chatbots that just came out in the last couple of years, whereas AI and machine learning have been around for decades now, literally.

When we talk about, for instance, agriculture, there’s probably very little or no generative AI involved. It’s really more the machine learning and making predictions and such, and looking at the correlations, what will give you a better crop yield, and such.

Farida: Right.

Karen: In cybersecurity, is that also true? Or do you see generative AI more creeping into that?

Farida: Yes, I’ve seen it in agriculture. Like in Canada there is a farm, because they deal with the loss of weather issue for the cold especially. They’re major exporter for the apple. So how to preserve the apple from being frozen or from the cold or things, how to make the size of the apple, the taste of the apples, because apples is huge varieties. AI can really help with the irrigation, with how to protect it from the weather and even with the climate, how to put the shades, the irrigation process, supply chain. It really, really helps and saves us a lot, and reduces the waste, which is something for the CO2 and climate change as well. It can really help. So it works. I have seen lots of programs implemented in many rural areas, in many, the certifications in third world country. I don’t want to label names, so people don’t get into this, but there is undeveloped world that they are not aware. I hope anyone, any woman who are aware of AI technology, to introduce it to these people, to farmers, because it really works.

While in cybersecurity, still we need to advocate, especially if you have children, if you are dealing with your personal life. I know people that they get an anxiety when people know too much about their life, their health. They need to be safe and protected, and their tolerance is limited. We need to respect that because there is no one solution for everyone. We need to accept the differences and limitation of each one of us.

Karen: Yeah, that’s a good summary, so thank you for sharing that. I’d like to hear a little bit about your writing process for writing these two chapters and what role AI tools played, if they played a role in your writing.

Farida: First of all, I’m an old-fashioned, still. I’m still analog, so I still write notes. So I’ve taken a lot of writing workshops, and in my writing workshop, I remember that we need to make a skeleton. So I first list my ideas, and once I list, I’m scaffolding them. So one, build on another, to have fluidity, to keep the reader engaged. And after that, I need to do the due diligence, which we call the arms and legs, how we are supporting these messages. And actually, this is where AI comes in because sometimes, especially in fields, like, there is a daily research or weekly research or new updates, I might be outdated with my information or I’m not aware of what’s happening. And here AI comes handy when I tell him, it’s like, “What is the last research on this one? Check my data and verify me some citations that cement or enhance my argument or to give this credibility or trust.” And AI, I stand them in this one perfectly well, better than me. I’ll never be aware of what’s happening because it can scrape all the internet faster than me.

And, as well, because English is not my first language. I assume I master it, but it’s still some metaphor and some little things that can be lagging behind and out of translations, like the sentence structure. Grammar-wise, he corrects me. But my notes here come handy because I make sure that he doesn’t change my tone, or the message I want to deliver. Because, you know, every time with every prompt, he adds a layer, a layer, a layer, and then tweak it, tweak, it, tweak. It’s out of context. It is not what you want. It’s no longer your voice. It’s related, but no longer your voice, no longer your message. So here, my notes — I still write with pen and paper, and it’s really helped me to stay authentic. I’m with total agency. This is my work. And when I go in a meeting, I know what I wrote. I know what is the output. I can debate it freely because it’s my angle.

So AI helped me, yes it does, in tightening my verbose, to be more accessible. Part of my job is writing reports. I know that some reports need to be brief, so I know every word has a weight. It’s not about the lens. It’s not about how information you give. It’s about how precise and concise you are. You need to, not strip it from emotion, but not to strip it from the value or the message it has to deliver. And there is psychological and strategy, and AI can help you with that, because sometimes you want to share too much, but sometimes it’s not needed. It can take you off-road. So AI can really help if you are in control. If you know what you want, you have an image of the output desired, not to leave it to him. So I use it. I’m not going to deny it, but I use it often. I try to exploit it the way he exploited my data.

Karen: It sounds like you mostly use it for a search assistant and for editing. It doesn’t write for you. You write with it and have it help you.

Farida: If it writes for me, what my input is? Like, I don’t want to be reduced, to just to be a cog or a component in AI. I want it the opposite. AI is a tool in my component output.

Karen: Great. Well, I’ve read through your chapters, and I was really impressed. And I think readers will get a lot from it. I’m wondering, what do you think readers should expect, and what do you hope they’ll take away from reading your chapters?

Farida: What I hope always with my writing is I’m not making decision for others. I’m just giving them the awareness or the understanding how the dynamic and the complexity is, and they decide for themself. Because sometimes too much information is hurtful, and less information is hurtful. So I’m trying to give them what they need, the basic pillars to make better decisions. There is no right or wrong. There is no good or bad, but there is what fits you right now. What is good for you right now, what you are dealing with, you are the person. You need to be aware of yourself and what you need and what you can handle, what you can tolerate, because an anxiety is real. Our burning is real, formal is real. And you know what? Attention economy and exploitation economy manifest on that. So I’m going to help you to make a better decision, so you are going to have an equilibrium peace, you’re going to be a good person.

And this is what we want to do. We want to live in a society that we are all happy, or at least in peace. We are not in conflict zone, we are not in an anxiety, we are not under stress, because the information exists. And you know, as God say, “Seek and you find.” So if you want to seek something, I will share it with you. I’m not going to hold it. If you have a question, raise it. And here it is. If you are looking for something, whether is online, you should find it.

So in my writing, I try to demystify or simplify things for accessibility. This is what we can do. We cannot make anything choosing for people, because we are stripping them their growth experience, because you grow only by mistakes. So let them do mistake and learn for themself, but show them the route. Give them a guideline. Give them a framework and let them learn to maneuver. Like driving. You learn driving, you take driving classes, but the street will teach you something else. And your experience will strengthen you and make you like you drive looking at mirror or don’t looking at mirror, or like with the speed, with no speed, you handle your thinking, you’re talking on the phone. I know women who can put makeup while drivers, they do a lot of things, but this multitasking comes from experience. Yet you can handle the car. And you know how to maneuver. And this comes with practice, not by telling you left and right and rules.

So I hope to share awareness, make the information accessible so people can make decision for themselves. Because our brain function, we are born as well, not just with defective processing, or our processing takes time. We learn things this time and people don’t let time play its role. They want it now, now, now, now, now, like hit and run, let’s do it, and things like this. And haste bring waste. As Sam, our friend who has beautiful newsletter, we need to slow down. We need to digest, we need to conceive information, and we understand that nothing is standalone. It’s a cycle and everything is interconnected with each other. Although it looks independent, but the interconnectedness exists, and the fact and ramification implication exists. Just to see them or to be aware of them is actually what I want to give to any reader, to my newsletter or my chapter or anything.

Karen: Thank you for sharing all that perspective and your context. I think those are good lessons and I hope a lot of people will have a chance to enjoy your chapters! Is there anything that you were hoping I would ask you that I didn’t ask?

Farida: I hope someone to ask me, “What do you suggest, especially for my friends who have teenagers or children, what they should do with AI and privacy?” Because I have seen cases, especially with deepfake, and it’s unfortunately increasing. There’s people that they are, if you’re wealthy and you have a daughter, even although you are keeping it herself, they’re going to take a picture and make a movie of her and still, starting extortion.

So for this reason, I keep saying: Protect yourself. Don’t share where you are. Or keep things private. My best advice for anyone, personal or professional to my client or to my friends: if you really worry about safety or privacy, keep it analog. Keep it out of the cloud. Keep it out of grid. So we need to consider the analogous process as a choice. It doesn’t mean that we are into innovative, to follow the technology, to follow the hype, to join the wagon, to do everything. Still, don’t overlook the past, because the past worked and worked really good. So keep it handy. Don’t just turn the page; keep it as an option. If you feel just to remove an anxiety from you, if it’s going to cause you a problem, if it’s going to affect your freedom, it’s going to affect, put you under stress or any peril – keep analogous as option. And this is actually my advice.

And there is a tendency now, even, for companies to start getting things, especially with the rise of agents, off the grid to keep it safe and protected and limit the access to the cloud. And we are going to see this increasing, I guess.

Karen: I’m glad you brought that up. We have a total of 24 chapters in the book. You’ve got two of them. But we do have chapters on parenting in the age of AI, and we’ve got others on education and how children’s data is being used. It’s definitely a really important area.

Farida: Yes. Yes. I interviewed who wrote about AI and parenting. And she was eye-opening, because I’m dealing from data perspective and she’s dealing from empathy and children and raising and growing. I don’t have children myself. And it was fantastic interview. I need just to polish it and publish it. She has lots of great insight. I learned from her perspective on how to shelter and grow in term and not to be outcasted. Because this is very important. You’ll see child, they are a mimicry society, and they go to kindergarten or go school and everyone with pads and telephones, and why not me? So we need a collective will. Every society needs to decide the collective will on how to protect children, especially parenting in certain ways. It cannot happen in isolation.

Karen: Thank you for sharing all that!

Farida: No, thank you so much for your time and for the opportunity to talk and I’m happy to meet you actually in person, virtually.

Karen: Likewise!

Farida: It’s, yes, we have been talking a lot and here I am. I’m ready to answer if anyone have question regarding my chapter or any verse or information or explanation, I would be happy to tackle them one by one.

Karen: Okay! And if you can just say the name of your Substack newsletter for everybody?

Farida: Yes. It’s Lights On because my job is to shed light: “Okay, look at this. This shadow exists. But if I put lights on it, maybe you take account for it.” Yes. So it’s Lights On. And I write about economics, business and some behavior, how we interact with things.

Karen: Great. Well, thank you so much. Your turn, I think!

Lights On by Farida

A systems-driven newsletter from a data engineer and cybersecurity thinker, decoding complex data, technology, and business trends into actionable insight.

By Farida Khalaf

Farida Khalaf interviews Karen Smiley

Farida: Yes. I would like to hear about, I know women can multitask, but how can you shuffle the challenge to deal with diversity? We are in different locations geographically, different time zones. What was the challenge and struggle you faced in collecting 24 women to work together in one project? Because I understand as is it, I work as team managers, like how to handle them, but sometimes it’s difficult to put them in a tandem.

Karen: I’ve been working with global teams for over 20 years. Even back in the early 2000s, I was working with a global team of researchers. And I ran projects with people that I never met in person, mixed teams across different countries and different types of experience. So it’s something that I’ve gotten comfortable with. And I really enjoy meeting people from other cultures and learning their perspectives. That’s something that I welcomed. And it’s also something that I think made me aware that we don’t all share the same values. And that’s not a criticism that any one set of values is better than the other, but they’re just different; different communities. And I learned about this from as well — her chapter on the African philosophies and they think much more of community, as opposed to this individualistic view that is so prominent in the US. And it’s, I think, just so interesting to learn about that and to challenge my own assumptions.

Everyone that I’ve met in other countries, I’ve learned something from. That’s certainly been true on this book. And that was something that I felt was important to highlight in the book, is that we have all of these different perspectives and experiences, and for the most part, these voices have not been heard. They’ve been drowned out by the people with all the money and the power.

Farida: Yeah.

Karen: And their hype just drowns out the voices of real people who are doing real things with AI, with generative AI and non-generative AI. And their voices and their experiences, and the biases in the data and in the systems that are being developed, they were just kind of hidden. And that’s why I ended up starting my own podcast originally in August 2024, to try to highlight some of those voices.

And then, I think you know the origin story for – about how we saw this post on Substack for someone saying, “Here are 10 people you should be following on AI”, and all 10 were men. And I thought, you know, this is just not right. So I messaged the guy privately and I said, “Look, there are no women in your list”. He said, “Oh, I just didn’t know of any.” I had only been on there for some months and I was already subscribed to 34 women, so there’s no shortage of women. And so I said, “Oh, okay, you know what? No excuses. Here’s my starter list.”

And then I started finding more, and then other people started telling me about more. And from December 2024, it grew from this list of 34 women that I had already found. We’re recording this in early March 2026. We’re well over 600 women from more than 60 countries. That diversity in so many dimensions is out there. We have people across 16 different categories. And that diversity and that set of perspectives, I learn from every one of them. It’s a lot of work, but it’s really been a joy to discover, and especially to find newer writers that are just getting started and to help them build their voices and help them connect with their audiences. And I’ve just really been loving that. That’s where that came from. So wrangling 24 chapter authors for a book is — again, it’s some work, but it’s so rewarding. I just loved it.

Farida: Yeah. I’m sure it was challenging and joyful in the same time because lifting other women, for me, it’s a pleasure. Especially people who comes from my culture from the eastern hemisphere. We were raised to be silenced, to be set aside. And standing up like really, it’s something we enjoy. Yes, we really enjoy.

So, you wrote a book about AI ethics and I got it. It’s very informative and very eye-opening in many avenues. And then you make the AI Everywhere. Is there is any connection? Or you have this vision to scaffold? Is the idea, “Let’s build the ethics awareness; now let’s get the domain how to apply it?” Is there a connection between them, or are they totally independent for each other?

Karen: Yeah, I think they’re more closely related than even I realized at first. Because this whole mission of supporting women, supporting people who’ve been underrepresented and overlooked and don’t have the voice that they ought to be able to have, that was the impetus for the Everyday Ethical AI book, to try to help people learn about the risks that were out there. Risks of privacy, and as you mentioned, the deepfakes and just everything that goes with that, the environmental impacts and the exploitative practices that underlie it, that most people weren’t even aware of. And the choices that we aren’t always given or that aren’t being made clear to us. So that, that whole aspect of it is why I wrote the Everyday Ethical AI book. I felt like it was a way that I could help more people to learn about that, than I could reach just through my own newsletter directly.

And at the same time that I was launching (that directory) in response to that post, I was invited by a professional publisher to be in a collaborative book as a chapter author. And my chapter was going to be on AI and data for entrepreneurs. And I thought that this project would help me to get established as a published book author and learn the ropes of publishing. So I invested in it. Now, that project has been delayed and that book is still in the works. At least I think it’s still in the works as of now. When I wanted to publish my own book, I decided I’m just going to go ahead and do it myself, and I’ll teach myself the ropes as I go. But that project kind of planted the idea of a collaborative book in my head.

And then I was starting to look in our She Writes AI directory. I said, “Well, let me find other women who have actually written books and see how many.” We had hundreds of women, and even now we only have, I think, about 40 books in our book list, which seems awfully low. So I started thinking, “Why haven’t more women published books?” I mean, Substack’s a platform for writers! You would think there would be lots of books and lots of writers, but I wasn’t finding them. So that seemed quite low, and I started thinking about why.

Then as I was working on my book, I’m realizing, “Okay, this self-publishing, this is a lot of work.” And there’s a lot of very steep learning curves on the tools and the systems, and the process you go through with getting advance readers, and just everything. Everything outside of the actual writing probably took more time than my actually writing the book. And that could be, I thought, something that would inhibit women from publishing.

Even just to be a contributor on the professionally-published book is expensive. And publishing a whole book by yourself can be expensive. So I thought that cost could obviously be a factor. And then as I got through learning how to publish my book, I thought, “Well, maybe this is something else I can do that would help more women to build their visibility, to build their authority in their fields, to be able to participate in a collaborative book.”

And that was really the impetus for the idea. So I made that post and put out the invitation. And I got such great response. I was like, “Wow, this is something that we can do.” I initially thought there would be some money involved because of outlays for things like cover design and editing and the interior book design. Some of the authors said, “You know, I’d love to do this, but that’s a lot of money.” And so they withdrew.

But we were lucky. We found some additional volunteers besides me who have experience in publishing some of their own books. So with that volunteer effort, and we got one sponsor who was very generous – among all that, we were able to make it so that the authors didn’t have to put in any cost share on this book.

Farida: Yeah.

Karen: So there’s no cost required for any of the chapter authors. And I’m hoping that we’ll be able to keep doing that for the future volumes. That, I think, will take away another barrier for some of the authors who weren’t able to participate in Volume 1.

And I decided that it needs to be a series. 24 sounds like a lot of authors and topics, but AI is so big and our world is so big, and there are so many experiences and voices that need to be heard. It’s like, “Nope, we’re doing Volume 2.” And then maybe Volume 3, we’ll see.

Farida: Yes, like the AI agent, it’ll need an update in maybe three months or six months to add what’s happening, because it can be outmatched for what is currently doing it right now. I learned from you actually, because I published a book about Women to Remember. I selected 100 women that already shaped our society in terms of art, technology, science activism. Their voices, I wanted to shed light on them. They exist and we can pursue them, not to be silent or step aside or to have a coffee. Women already tackled them, every avenue possible. They challenged everything in front of them. All the struggles they dealt with, and some accomplishments need to be cherished and actually spoken out about it out loud. And teach our children, “Don’t be afraid of challenge. There is a woman who dealt with that. Check her out. And learn from her and build on it. Use it as a wrapper.” There is always a woman who dealt with that. Regardless where you are or your education or your culture, there is a woman who did it, who went through what you are dealing with right now, from 1500, or 500 BC, they existed.

And we need to believe more in ourselves. The book is more about woman empowerment, but it aligned with you from the angle, like we should stand and we should let our voice heard and be known. Our heroes shouldn’t be the Avengers or, you know, Catwoman, or Hercules or all these male figures. We have women who shaped history, who shaped science, who shaped technology, who shaped the legal arena, who shaped everything, who shaped civilization itself in leadership. And we should take heed of them.

So I love your AI community that you are building because I can see that – I don’t know, what is your metrics? – it’s like I know that everyone is welcome. And your guideline has all to respect each other. And all the people engaging, actually, they’re extremely professional. Our exchange is really, really helpful. I don’t know if you have a metric or guidelines for acceptance, but it is the fact that it’s open, that it’s not biased toward any mold to fit in, which is actually the beauty of it. Because anyone can be, just be herself. Come with your culture, come with your knowledge, come with your experience, and just share it and let’s grow together.

Karen: Yes, that’s exactly the spirit of the community. The requirement for joining is: if you’re a woman or non-binary, then you just fill in the form with the information that you want in your directory listing. Pretty much the only person – I’ve left them in the directory, but I took out their newsletter – was a newsletter that wasn’t suitable for people that are underage. Because we do have some high school students who are in our community, and they’re listening. And I wanted to keep the digests safe for all ages. So that one, she’s in the directory, but her newsletter doesn’t go in the digest. But other than that, pretty much, it’s open to any woman, cis or transgender, it doesn’t matter. And I wanted it to feel welcoming and supportive. And I hope that vibe has come through.

Now for choosing people for the book, that was a little different. Because there we’re looking for people that understand, have authority and have knowledge in specific topics, and have experiences that they can share, and that they are comfortable writing about and writing well. We got some great responses. And I’ve actually got a lot of good expressions of interest on Volume 2 already. So that’s really proving to be fun.

But for the book. It’s funny, on Substack the other day, someone asked me, “Well, is this an anti-AI book?” I think the way she put it was, “I hope this is an anti-AI book?” And my answer was, it’s not. It’s not. The book isn’t anti-AI, and it’s not pro-AI. Our community has a whole spectrum. The requirement for the community is not that you use AI, but that you write about it. And some people write about what their concerns are, their ethical concerns or practical concerns and why they don’t use it, and that’s okay. And there are other people that write about why they do use it, and that’s okay. We’re not here to judge people for how or why they use it. We’re here to help support writers in writing about AI. The focus is the writing.

So the book, I would say, is neither. It’s not pro-AI; it’s not anti-AI. We’ve got kind of a mixture. And that’s intentional, because that’s what reality looks like. Reality is not on either of those extremes.

Farida: I read the book because we were in the peer review process. I have to thank for doing a great job with me, and you as well, with your patience with my chapter. It was a learning curve actually. The fact that everyone sharing his perspective, you cannot say it’s, “We are going this way.” It’s like a point and the vector, and everyone is coming from a vector perspective, from a different angle, which is actually a bit of richness.

People will enjoy it because you will focus on the chapter or the domain you are interested in, and you are going to read a professional perspective. I know that the launching is hopefully on 15, it’s already on Amazon. Congratulations! It’s on the top one tier suggestions. I already published it on LinkedIn as well, because it’s an accomplishment.

So what do you hope for Volume 2? What is going to be Volume 2? Is it a continuation of Volume 1, or it’s going to take a different angle?

Karen: Yeah, it’s a collaborative work. My overall vision is that we have books that share experiences and practical tips and diverse views and values that people have around AI around the world. We have some rough categories. We have home, society, and business and the how-to’s. But beyond that, I just want to see what people want to write about. I didn’t try to control that in Volume 1, and I think it turned out pretty well. We got a really good distribution. The only thing that I really look for is to make sure that we don’t have two people writing about pretty much the same thing. Each chapter is distinct and adds something different that’s of value to a reader.

Beyond that, I’m excited to see what proposals come in. I’ve approached a few people. I did go out of my way to look for people who are outside the US, so we don’t dominate the book. I think that’s important to have that diversity aspect. And if it doesn’t arise naturally, then I’m going to go look for it. I think we’ve got a good balance right now on the list of proposals that have come up so far for Volume 2. And probably some of them will shift out to a Volume 3 that would be next year. But I like the momentum that we’re seeing and the interest.

Volume 1, it’s a #1 Hot New Release. We haven’t reached #1 Bestseller yet, but I’m really hoping that we get there by launch day on Sunday. It’s always category-specific. It’s like, we’re not the #1 Hot New Release in all of Amazon, but within our categories, right? Yes, that’s pretty cool. We’re the #1 Hot New Release in Social Aspects of the Internet, and that’s the category that Cory Doctorow’s ‘Enshittification’ book is in. So that’s a pretty good category.

So it’s exciting to see that people are already recognizing, “Hey, this could be a really cool resource for me to learn and to understand.” I’m excited about that. And I hope that we can connect with more people. That was the whole goal of the book, for us to connect among ourselves as writers, and for us to reach readers that we don’t reach otherwise, who maybe don’t live in the Substack ecosystem like we do.

Farida: It’s very important, the visibility. Because actually, the time we’re living in, we need to be visible. And the only thing to make us visible is our work. Because in the abundance of everything, originality and genuine voices is actually the way to stand out. So for the AI community, I know that you write the curated digest list. I’m happy to be featured in it often, but this curated list is actually a good process for people to start with.

So do you have any vision for the She Writes AI community? Not just for the book, how to continue in the progress. Are you hoping to create like a library on topics or domains so people can go to it? Or to be Women Wikipedia for AI? I had everyone share his perspective, any plan with the community, because it’s actually a great initiative.

Karen: I like that idea. Women’s Wikipedia for AI – we might have to put that on our idea list. At this point I think we’ve got even more ideas than we have volunteers. We’re all volunteers, so we have to work within, everybody’s got lives. Women especially tend to – they’re basically working two jobs, right? Work and at home. We all have interest and passion for this, but we also have limitations.

Farida: Yes, yes.

Karen: So we have a lot of ideas on things we’d like to do. The digests were the first thing, and every week I hear from more and more people about how valuable it is to have this one summary of over 300 articles from our community in different categories and highlighting some wild cards that might be overlooked. And people are just really finding that valuable. And I am posting them. I started up a community presence on LinkedIn for us to help to reach, again, new people that we aren’t connecting with necessarily on Substack. And that seems to be working well. So I’m going to try to continue that. The tool is pretty automated, thanks to some work that had done last year. There’s still some manual work, but I’m hoping that I can keep that going.

And obviously the books, once Volume 1 is out. I do have my own second book on ethical AI that’s in the works, but I’ve kind of had it on the back burner because I really want to get our book series established and out there. I’ll personally come back to that. But I really feel like I’m enjoying so much our She Writes AI community, what we’re doing and how we’re supporting each other. I just love seeing that. And we’ll look at what the community wants and what people need and what fits with within their bandwidth.

One idea we’ve kicked around is starting up a mentorship program. Especially for newer writers or for people in countries where they haven’t had that support system so much. But again, we’re all volunteers. The mentors would be volunteers.

Farida: Yes.

Karen: And so we don’t want to launch it and then not have the time put into it to make it successful and make it work for people. So I think when we get more volunteers, or if someone wants to step up and lead that, then we can look at launching it.

I need to go find your book, Women to Remember. Have you seen ‘s new Herstory series?

Farida: Yes.

Karen: Where she’s profiling the women in AI. That’s a cool connection there.

Patrick had set up our first networking session. We had that kicked off for the first time last month in February. That was, I think, pretty successful. We made some good connections there. I’d like to see more of those events. As I said, we’re getting more volunteers. Cristina had volunteered to try to help coordinate those events, but I’m sure she would appreciate having more people involved. Live events could be interesting. I don’t do video myself. I promised my family that, just as a matter of privacy. So I won’t be doing recorded video, but anyone who wants to, I think that’s totally cool.

Farida: Yeah.

Karen: If we want to start up a podcast, maybe initially for these author interviews, then we could also end up using a She Writes AI podcast for other interviews. Maybe interviewing with a community member and just have them talk about what they do with AI. And that would be another way that we could help people get visibility. You could host it, if you would be interested. Something to think about. :)

Farida: Yeah. Yeah. I think with the topic I’m interested in, like with Lakshmi, with Dhani, lots of topic is overlapping, I’m happy to discuss it. Because I have some, I’m not going to say it’s the same depth as them, but at least the understanding to handle the conversations. From the other one, I’m just a reader. It depends on which chapter. I’m not going to say it’s like I wear different hats, but it’s still in certain domains. Like this one, I can converse properly and freely and can add a value or get a value. To the other, I’m just a reader who learns.

So I can participate for sure. If there’s anything from my end, I will definitely participate in the community. In woman empowerment. Women solidarity. Count me in. Immediately I would say yes, yes, yes, 100% yes. Because living and traveling, in many even rural area, I realized that we need to reach out, we need to destroy these walls. We built social images, social criteria. And it built walls and left many people outside. And I think we need to destroy that by just tearing this wall down and reaching out: “Okay, you are very welcome.”

And actually, this is how we are going to get stereotypes. If you are this, you are that. I was labeled everything: being Egyptian, being Muslim, being terrorists, being bully, being outdated, being conservative. We heard it all and we dealt with it. And stereotype is real. It’s not just bias. We need to give benefit of that and reach out to have a better tomorrow for our kids and for our society. We don’t need to be stressed and, as God say, “Seek and you will find”. I believe anyone knocks your door, you should answer as long it’s coming with good faith. And I think your initiative is really, really, really helpful. And it’ll help a lot, especially to raise people, sound, voice from Africa, from Asia, from Latin America, different voices, different perspective, different culture is amazing. It is really amazing. I’m really impressed and I’m happy to be part of it.

Karen: Yeah, it may have started as my community, but I think it has become our community, and that just makes me really happy. So yeah, I’m glad you’re part of it!

Farida: Me too. I really enjoyed being welcome in your group. Thank you for the opportunity to write these two chapters. I hope we will enjoy them.

Karen: Thank you, Farida. I appreciate your time and thank you for interviewing me! And we’ll look forward to hearing your other interviews with the other authors. I can’t wait.

Farida: Yes!

6 'P's in AI Pods (AI6P)

AI affects People & Places we care about, Products & Platforms we build & use, and Practices & Processes we use daily. Wondering how data & AI impact these 6 'P's, and how to use them ethically & wisely? The AI6P newsletter is for you! 100% human-written.

By Karen Smiley

References

Announcement

Everyday Ethical AI

📢 Announcement: Volume 1 in the She Writes AI Everywhere book series is here!

Hi folks! I am excited to share some big news with you all. You may have noticed that I’ve been writing less often in this newsletter lately. I’ve been happily busy curating and preparing to publish our forthcoming SheWritesAI book “AI Everywhere, Volume 1: How Women Are Changing The World With Artificial Intelligence…

Read more

2 months ago · 30 likes · 2 comments · Karen Smiley

Book purchasing information

• Buy the ebook on Amazon (now also available in paperback and hardcover)

• Tips from our publisher on how to buy an Amazon ebook from a non-KDP-marketplace country, and how to read it without a Kindle device

• Purchasing and review links (we’re on Amazon, Goodreads, and Storygraph)

aiEverywhereBooks.com - our beautiful book series website created by

She Writes AI Community book series info