Rebecca Mbaya & Farida Khalaf: Author Interview #22, AI Everywhere, Volume 1
AI Everywhere, Vol. 1, Ch. 2: AI and African Epistemologies: Why communal wisdom is the antidote to data colonialism & how builders in the Global South are operationalizing "Sovereignty-Before-Speed"
Welcome back to the She Writes AI interview series, where we talk with the authors behind AI Everywhere, Volume 1, exploring their personal journeys and professional expertise to understand how we can build a future for AI that is truly ethical, responsible, and human-centered.
This is the 22nd episode in our series, and today I am joined by Rebecca Mbaya. Representing the Society section of our book, Rebecca’s Chapter 2: AI and African Epistemologies provides a profound critique of “data colonialism” and the narrow Western worldview that currently dominates AI development. As a Congolese social innovator based in South Africa, Rebecca founded The African Innovators Series (TAIS) to spotlight changemakers who are building technology under conditions of constraint—revealing that what Westerners see as “anomalies” are actually the global baseline for most of humanity.
Rebecca Mbaya brings a unique perspective on Theory-Bearing Practice to this conversation. She argues that African philosophical traditions like Ubuntu (”I am because we are”) and Yoruba contextual wisdom are not just abstract concepts, but operational strategies used by innovators to survive and flourish. She moves the discussion beyond “catching up” to the West and instead explores how we can reorganize how intelligence itself is conceived—shifting from individual models to collective intelligence.
In this final conversation, Rebecca explores three practitioner-named frameworks: the Ecosystem Tax, Sovereignty-Before-Speed, and the Warmware Layer. She discusses why we must anchor data and value locally, why relationships are a form of infrastructure, and how building AI for reality rather than idealized environments is the only path to a truly inclusive digital future. Whether you are a global funder, a developer, or a community leader, Rebecca’s insights offer a roadmap for building technology that respects human dignity across every border.
Below is Rebecca Mbaya’s full interview response to Farida Khalaf’s questions on AI and African Epistemologies from AI Everywhere, Volume 1.
Topic: AI and African Epistemologies (Chapter 2)
Intended Audience: For social innovators, global tech leaders, policy makers, and ethicists seeking a disciplined framework to move beyond Western-centric AI narratives and understand how communal, contextual, and relational logics can build more durable technology in resource-constrained or trust-fragmented environments
Part 1: The Professional Path & The Collective Mission
The Expert Perspective": Rebecca Mbaya, you are a social innovator who launched The African Innovators Series (TAIS) in March 2025 to spotlight perspectives that rarely reach global audiences,. What was the spark that made you realize that the dominant AI narrative wasn’t just under-representing Africa, but was actually suffering from a “misalignment at the level of design”?
Rebecca: Honestly, there was no single spark, because the misalignment was never really a surprise to me. The dominant narrative about Africa has always suffered from misalignment at the level of design. Those who control the narrative have never designed it with the goal of presenting Africa from a place of authority. Under-representation is the intended outcome. So, what is happening with the AI narrative is not new.
What TAIS did was confirm both things in real time: that the misalignment exists in AI just as it has everywhere else, and that African innovators are operating from a fundamentally different theory of knowledge despite the constraints. The series began as visibility work, but over eight months and 36 conversations (at the time of writing), the patterns that emerged made it impossible to ignore. Our theory of knowledge shapes practice under constraint and the constraints we have been forced into, as our history makes plain, did not happen by accident. When you examine contemporary AI’s embedded assumptions against what African innovators actually do, the gap becomes undeniable. The way AI is built today: centering individual cognition, treating data as something to extract, prioritizing speed above all, is not how African innovators naturally approach the work. And yet these assumptions are rarely questioned because they are presented as the only logical way to build. They are not. They are philosophical commitments from specific traditions that got to define what “normal” looks like.
The Global Tribe: As one of the authors in this collective representing five continents, your chapter challenges the idea of data colonialism. How does it feel to provide a platform for 34 innovators whose work proves that intelligence is not something to be extracted, but something that emerges through community?
Rebecca: It feels like responsibility more than anything else. I believe that every African at this point should be aware of the responsibility they carry in this moment; at whatever level we are at, there is something that can be done. As little as it may seem, it still counts.
The innovators I feature are doing their part, most of the time without the support they need and deserve. And yet they keep building, and they keep centering their communities, because they understand the responsibility they have too. And when you look closely at how they work, that sense of responsibility to community is not just a value they hold, it is how they build. It shows up in every decision, every framework, every solution. That is what proves that intelligence is not something to extract, it is something that emerges through people, through relationships, through showing up for one another. So providing this platform is my part of that same responsibility. And the chapter is my attempt to make sure that what they are doing is not just seen, but understood for what it really is.
Relationships are not an adjacent system; they are the medium through which technology becomes useful, credible, or meaningful.
Part 2: The Operational Frameworks - Ecosystem Tax & Sovereignty
The Ecosystem Tax: You cite McKevin Ayaba’s term, the “Ecosystem Tax,” noting that African builders often spend 60% of their resources building infrastructure that should already exist. For a leader in a mature economy, why is it vital to understand that an African startup is often “5 to 7 businesses in one”?
Rebecca: Because if you don’t understand that, you will keep measuring African startups against a standard that was never designed for the context they operate in. A leader in a mature economy has never had to build the road and drive the car at the same time. The infrastructure they rely on was already there when they started. They inherited an ecosystem that works. African builders often don’t have that luxury.
When McKevin Ayaba coined the term Ecosystem Tax, he was naming something every African builder knows viscerally. So when a context-driven African startup looks like it is moving slowly, or spreading itself too thin, or not scaling the way a Silicon Valley company would, that is why. They are not building one business. They are building five to seven simultaneously just to make the first one possible.
And yet many succeed. That is the part that rarely gets told. When an African startup succeeds despite the Ecosystem Tax, it is not just a business success. It is proof of an extraordinary level of resourcefulness, resilience, and capability that most leaders in mature economies will never be tested on. So for a funder, a partner, or a policymaker, understanding the Ecosystem Tax is not just about empathy, it is about accurately recognizing the value and the strength of what is being built.
Sovereignty-Before-Speed: Dr. Isaac Bayoh advocates for building slower to ensure data and value stay anchored locally. Why do you argue that Silicon Valley’s “move fast and break things” mantra can be a marketing ploy for extraction when applied to fragile or resource-constrained contexts?
Rebecca: Because speed without sovereignty consolidates extraction. “Move fast and break things” emerged in ecosystems with abundant capital, stable institutions, and safety nets. In those contexts, breaking things is a calculated risk; there are systems in place to absorb the damage and course-correct. But in fragile or resource-constrained contexts, breaking things produces harm that consolidates outside power rather than catalyzing innovation. Value moves outward, control disappears, and communities are left dependent on external infrastructure that can shift, restrict, or withdraw access at any moment. And in fact, there is not much left to break. Whatever could be broken has already been broken throughout our history.
That is why Dr. Bayoh’s position is so important: build slower, but ensure that data, models, and value stay anchored locally. Because Sovereignty-Before-Speed is not about rejecting technology or ambition, it is about understanding that we don’t have the same realities. Historically, it was the speed at which outside forces moved that cost African communities their sovereignty in the first place. That pattern did not end; it evolved. Sovereignty-Before-Speed reminds us that long-term strategic strength belongs to those who retain control, even when the early pace appears slower.
Contextual Fidelity: You mention that Yoruba philosophy emphasizes contextual wisdom over abstract rules. How does this translate into a Sovereignty-Before-Speed framework where a model must prioritize local languages and legal frameworks before attempting “universal” scale?
Rebecca: Yoruba philosophy teaches that truth holds within circumstances, not above them. There is no universal vantage point from which to build something that works for everyone. And yet when it comes to Africa, context is rarely treated as the foundation, the infrastructure on which everything else is built. Instead, it is reduced to a variable to be consulted, an afterthought to be accommodated. That is the misalignment. Context is not a variable, it is the infrastructure, and that is exactly what Sovereignty-Before-Speed in Africa operationalizes.
When you try to scale universally before getting the context right, you flatten the very complexity that makes a solution legitimate. A legal AI model trained on Nigerian statutes cannot substitute for Kenyan legal dynamics shaped by distinct colonial histories and post-independence reforms. Swahili varies across regions; models that erase those differences may function technically, but they fail socially. That is not a minor gap. That is the difference between a tool that serves a community and one that is simply deployed on it.
What the innovators understand is that building from context first is not slower, it is more honest. In terms of African languages, for instance, translation-first approaches may seem faster, but they tether African languages to English-centric architectures. Building from scratch, investing in local datasets, and insisting that African languages stand as primary categories rather than peripheral add-ons: that is sovereignty in practice. Universal solutions flatten complexity for scale. Contextual solutions preserve complexity for legitimacy. Yoruba philosophy has always known this. Sovereignty-Before-Speed simply applies it.
Part 3: The Warmware Layer & Relational Design
Defining Warmware: Jason Bygate coined the term “Warmware Layer” to describe people, their motivations, and relationships as infrastructure. Why is it a mistake for tech teams to treat the “human substrate” as a soft afterthought, rather than a core system requirement?
Rebecca: Because you can build the most technically perfect system in the world and it will still fail if the people it’s meant to serve don’t trust it, don’t see themselves in it, or feel threatened by it. The technology is the easy part. The human layer is where things actually succeed or fall apart.
And in African contexts this is even more critical, because the people have a way of thinking and doing that is often not paid attention to in the design process. So when a solution fails, the blame is rarely placed on the design. It gets placed on the people. Digital illiteracy, resistance to change, lack of adoption: these become the explanation. But that is a deflection. The real question is why the human layer was never treated as infrastructure to begin with.
Jason Bygate’s concept of the Warmware Layer names this directly: people, their motivations, their relationships, and their capacity and lived experiences are not a variable to manage after the system is built. They are the foundation the system must be built on. When you treat warmware as an afterthought, you are not just making a design mistake. You are setting up the community to be blamed for a failure that was designed in from the start.
Relational Technology: Drawing on Akan philosophy, you suggest that technology does not meet “users,” but dense networks of hierarchy and trust. How does a “Warmware” approach change the way we evaluate whether a diagnostic AI system succeeds or fails in a clinical setting?
Rebecca: It changes what you measure. A diagnostic AI can be accurate, fast, and technically flawless and still fail, not because of anything wrong with the tool, but because of where it lands in the existing network of relationships, hierarchy, and trust in that clinical setting. Akan philosophy understands this deeply: personhood is not contained in individuals, it is constituted through relationships. So when a technology enters that space, it is not meeting users in the abstract. It is entering a dense web of meaning, power, and professional identity.
A Warmware approach asks different questions. Does this tool feel like a partner to the clinician or a threat? Does it stabilize existing relationships or destabilize them? Is it being voluntarily adopted or just deployed? Those are not soft questions, they are the real indicators of whether a system is working. Technical benchmarks tell you what the tool can do. Warmware metrics tell you whether it actually belongs in the space it was built for.
In a clinical setting specifically, a diagnostic AI that unsettles professional identity or shifts power in ways that were never addressed will face resistance that no amount of technical improvement will fix. The solution is to go back to the relational environment and designing the tool to function as a partner in human expertise, not an intruder into it.
Ubuntu has always known: under constraint, survival requires collective action.
Ubuntu in Action: Only two of your interviewees mentioned Ubuntu by name, yet all of them practiced it,. Can you explain how the “I am because we are” philosophy makes collective burden-sharing the most rational and efficient response to structural absence?
Rebecca: That gap between naming and practicing is an interesting finding in this research. The philosophy is present in every decision, every collaboration, every response to constraint without being cited. That tells you something important about how deeply rooted it is. It doesn’t need to be named to be operative.
And when you understand that, the answer to the question becomes clear. When the structures that should support you don’t exist, the only logical move is to build them together. That is not idealism, that is pragmatism. And that is exactly what Ubuntu operationalizes in practice.
In contexts where the “we” is still under construction; you cannot afford to wait for someone else to build it, and building it alone is how you pay the Ecosystem Tax. So collective burden-sharing is not a cultural preference in this context; it is the most efficient and responsible response available. When payment infrastructure is co-developed and shared, when talent pipelines are built through joint training, and when policy gaps prompt coordinated advocacy, the rising tide lifts all boats. That is “I am because we are” as an operational equation, not a philosophical ideal.
The contrast makes it clear. A competitive individualist lens rewards isolation. A centralizing tradition tells you to wait for the state. In the case of African innovators, Ubuntu makes collaboration feel like the most natural and rational path, because under these conditions, it is.
Build slower, but ensure data, models, and value stay anchored locally.
Part 4: Governance & The Global Baseline
Data Sovereignty: You argue that genuine consent requires the ability to withdraw, not just a one-time signature. What choices must developers make regarding GPU clusters and local compute capacity to ensure they aren’t consolidating extraction?
Rebecca: Genuine consent requires the ability to withdraw, to reclaim data, to adapt or stop a system independently. And that kind of consent is only possible when communities have actual control over the infrastructure their data lives on.
That is why the question of GPU clusters and local compute capacity is more of a sovereignty question than it is a technical preference. When all the foundational layers of an AI system sit on foreign cloud infrastructure, the community has no real leverage. They can agree to terms but they cannot meaningfully withdraw from them. Dr. Bayoh is clear on this: Africa needs GPU clusters within Africa. Not as an aspiration, but as a prerequisite for genuine control.
The choices developers must make follow from that. Build local compute capacity, even when cloud services are faster and cheaper. Train local talent, even when external expertise is more readily available. Develop local datasets, even when global corpora seem more efficient. Establish governance that is aligned with community norms, rather than imported templates. These are the conditions that make consent real. Without them, what gets called partnership is often just a more sophisticated form of extraction.
Challenging Universalism: Your analysis suggests that contemporary AI is not universal “human intelligence,” but a “historically specific way of organizing knowledge”. How can we begin to “unsettle” this model to better serve the “global baseline” of the world where institutions are fragile and resources are scarce?
Rebecca: We start by stopping the pretense that the current model is universal. Contemporary AI was built on a very specific set of assumptions that reflect the priorities and conditions of a small number of contexts. Presenting those assumptions as inevitable human intelligence is a political act.
The unsettling begins when we recognize that what gets called “fragile” or “resource-constrained” or “emerging” is not the exception, and it’s not the edge. It is the global baseline. Most of the world operates under conditions of scarce resources, uneven infrastructure, institutional fragility, and histories of extraction. These are not anomalies to be solved before the real work of AI can begin. They are the reality that AI needs to be built for.
That is what African innovators are demonstrating. Not an alternative paradigm for Africa, but a more honest paradigm for the world. The frameworks that emerge from contexts where constraints are sharper carry operational wisdom that is invisible in privileged settings precisely because those settings never had to develop it. Taking African practice seriously as theory-bearing is not about exceptionalism. We need to be asking the kind of questions that the dominant model has never had to answer: what would it mean to build AI not for idealized environments, but for reality?
Part 5: Actionable Awareness & Future Vision
The Substack Connection: Through your Substack and the African Innovators Series, you bridge the gap between “Northern epistemologies” and “contextual wisdom”. What is one “operational wisdom” from an African innovator you’ve interviewed recently that could help a Western developer avoid an “Ecosystem Tax” failure?
Rebecca: I’ll push back on the “one” because reducing 36 conversations to a single insight is exactly the kind of framing TAIS exists to challenge. The wisdom is not located in one innovator. It is collective, and that in itself is the first lesson.
But if I were to distill what came up repeatedly across the conversations, it would start with Michael Michie’s questions: before proposing any solution, ask why does this gap exist, and why has it remained unaddressed? Those two questions alone would save Western developers from importing solutions that were never designed for the context they are entering.
From there, Moses Maweu adds that technical excellence in constrained contexts means valuing simplicity over complexity: effective and sustainable with the resources available, not impressive by external standards. Khadija Iddrisu makes it concrete: design with low bandwidth, high cost, and constrained devices in mind from the start, not as an afterthought. Sion Israel Sion reminds us that systems must be built to work offline, intermittently, or through alternative channels. Merveille Mukoko shows that building trust and cultural intelligence is not slower, it is the work. And Rorisang Molefe makes clear that without the legal and governance scaffolding, even the best innovation will stall.
They all provide different layers of the operational wisdom. Understand the context before you propose the solution. Design for reality, not for ideal conditions. And never mistake the absence of infrastructure for the absence of intelligence.
The Final Message: You conclude with a powerful question: “What would it mean to build AI not for idealized environments, but for reality?” If a reader takes only one lesson from African epistemologies this week, what do you hope it is?
Rebecca: That epistemologies guide innovation even when they remain unnamed. The assumptions baked into how we build technology are not neutral: they come from somewhere, they serve someone, and they can be challenged. African innovators are already doing that, most of the time without the support, recognition, or resources they deserve. The lesson is not to admire that from a distance. It is to ask yourself what your part is, because there is always a part to play.
African AI practice is not an exception. It is a window into the conditions under which most of the world actually builds technology.
Why This Matters
Rebecca Mbaya’s work is a vital intervention because it unsettles the idea that there is a single, universal model of technological rationality.
The Ecosystem Tax: Rebecca highlights a structural reality where African builders spend roughly 60% of their resources constructing the infrastructure, payments, talent pipelines, and policy, that should already exist. This “tax” requires an Ubuntu-aligned response of collective burden-sharing.
Sovereignty-Before-Speed: She challenges the “move fast and break things” mantra, arguing that speed without sovereignty leads to extractive data flows. True innovation requires building slower to ensure that data, models, and economic value remain anchored in the local community.
The Warmware Layer: Rebecca reframes the “human substrate” as architecture. She argues that technical capability is rarely the obstacle; success depends on the “warmware”, the motivations, relationships, and trust of the people who determine if a system survives contact with the real world.
Contextual Fidelity: Her framework demands that systems align with local histories and languages rather than relying on “translation-first” approaches that tether African cultures to English-centric architectures.
Thank you, Rebecca Mbaya, for this incredibly grounding and visionary conclusion to our series for Volume 1. For readers and builders ready to move from “idealized” systems to those built for reality, her chapter in AI Everywhere, Volume 1 is a masterclass in how philosophy becomes practice.
Don’t miss her ongoing work in The African Innovators Series (TAIS) on her Substack, where she continues to spotlight the changemakers shaping the continent’s tech landscape. Rebecca’s final message is a challenge to us all: since most of humanity lives under constraint, we must stop asking how the world can catch up to AI, and start asking how AI can finally catch up to the world. I personally highly recommend her “Warmware” approach as the essential foundation for any project that aims to be truly human-centered.
And if you’re interested in more conversations like this, consider subscribing to She Writes AI. I’ll continue sharing insights from the voices shaping the future of AI, including contributors from AI Everywhere, Volume 2, where 30 women and nonbinary authors from across five continents continue to explore how artificial intelligence is changing the world we live in, and how we can change it for the better.
References
Book purchasing information
Buy the book (now on Amazon, Gumroad, Lulu, Ingram Spark, and many ebook platforms). Look for our upcoming abridged audiobook edition in July!
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)
Review links (we’re on Amazon, Goodreads, and Storygraph)
For more insights on the book series, 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:
 | A guest post by
|