These reflections began at the HealthAI Global Governance Forum, held in Nairobi on 2 December 2025 specifically during a session on the “Innovative Pre-Market Approaches to Validate and Regulate AI in Health.”
It was a session about the future of Software as a Medical Device (SaMD), regulatory adaptation, living labs, cognitive trust, continuous monitoring, and the daunting question at the center of it all: How do we regulate health AI fast enough to protect people – but not so slowly that innovation stalls?
If AI in health is alive – adaptive, learning, continuously evolving – then regulation cannot remain archival, document-driven, or episodic. It must become continuous, feedback-oriented, risk-responsive. And if the world’s leading regulatory environments are still figuring this out, countries like Kenya must think even harder about what it means to build, deploy, and govern AI at scale.
Because underneath the regulatory debate lies an even bigger one: Will Kenya become a user of AI, or a creator of it? And how do we avoid becoming another country stuck in what economists call the middle-technology trap?
The Middle-Income Trap – and Its Technological Twin
The middle-income trap describes economies that escape poverty, industrialize, then plateau. They become too expensive to compete with low-wage nations, but too shallow in innovation to compete with high-tech ones. They rise, then stall. Now comes the technological parallel:
The middle-technology trap happens when a nation adopts modern technology without building the capacity to produce it.
Sounds familiar, doesn’t it?
It happens quietly, especially in Healthcare. Health apps launch. Pilots succeed. But beneath the surface, infrastructure remains foreign, compute is rented, data governance is loose, regulatory science is embryonic, and sustainability is absent. Digital adoption advances – but technological sovereignty does not. Countries become digitally modern, yet technologically dependent.
Digital Health Has Lived in This Trap for 20 Years
The pattern is painfully familiar across our health systems:
- Innovation emerges.
- A pilot delivers results.
- Donor cycle ends.
- No scale, no continuity, no ownership.
- The system reverts to paper – or worse, to nothing.
Pilots are not progress. Platforms are progress. And yet, the former is what we reward, fund, publish, showcase, celebrate. The latter demands long-term ownership, domestic budgeting, regulation that learns, data infrastructure that doesn’t expire, and local engineering talent that builds, not integrates.
Most digital health solutions die not because they failed – but because we never built a home for success to live in. This is well understood by industry insiders.
Europe’s Struggle Shows the Stakes & Africa’s Risk Is Higher
Even Europe, with capital, compute, industry and universities, worries about losing technological competitiveness. Its challenge is the opposite of ours – regulation is heavy, adoption slow – but the destination is the same if action lags. If Europe fears stagnation, Africa must treat it as urgent. Europe risks over-regulation. Africa risks under-infrastructure. Different pathways. Same technological cliff.
Kenya today stands at a crossroads strikingly similar to emerging geopolitical middle powers such as Angola. As global power becomes multipolar, influenced by the US, China, EU, India, and new technology blocs, countries that are neither dominant nor peripheral must determine how they will shape their role.
Middle powers (Kenya is one) succeed not because they are large, but because they cultivate strategic autonomy, coordination capacity, and the ability to influence systems beyond their borders.
Kenya’s National AI Strategy (2025–2030) expresses precisely this ambition: to shift from being a passive consumer of imported technologies to being an active producer of AI models, data ecosystems, innovation, and governance frameworks. But middle powers fail when they modernise without building internal capability. That is the middle-technology trap: digitised but dependent, modern but hollow, advanced but not really sovereign.
The Trap Is Avoidable Because the Chasm Is Crossable
Geoffrey Moore’s Crossing the Chasm is often relegated to startup lore, a book founders quote, not a framework nations build from. But the insight applies at continental scale:
Innovation rarely dies because the technology fails. It dies because it never crosses into the mainstream.
That is the middle-technology trap in its purest form. We build pilots. They produce evidence. Evidence earns applause, but not integration. Nothing crosses. Kenya’s opportunity and risk is exactly this. Fail to cross the chasm, and we become an AI “user state.” Cross it, and Kenya becomes one of the few nations on the continent capable of defining, not merely absorbing, the future of health intelligence.
And unlike high-income countries, we (LMICs) are not entirely weighed down by legacy systems. We do not need to dismantle old EMRs, unwind vendor lock-ins, or rewrite 30 years of device-centric regulation. We are not late, we are structurally unencumbered. And that is a strategic advantage.
If African governments, innovators and regulators treat pilots not as endpoints but as prototypes for national platforms, we can cross the chasm faster.
| Pilot Thinking | Platform Thinking | Sovereignty Thinking |
|---|---|---|
| Donor-funded | State-budgeted | Market-driven |
| Tools | Systems | Infrastructure |
| Dashboards | Agents | Intelligence |
| Reporting | Learning loops | Self-evolving AI |
Pilots are how we try. Platforms are how we transform. Sovereignty is how we stay transformed. The trap only wins if we stop halfway through.
So, AI Could Either Break the Trap or Cement It Beyond Repair
AI is not just another tool, it is infrastructure. It can enable:
- adaptive clinical decision-support fine-tuned on local epidemiology,
- patient-voice monitoring loops that detect model drift in real time,
- health system agents that automate reporting, optimisation, triage,
- risk-based regulation grounded in evidence, not documents,
- local model development instead of prompt-wrapping foreign ones.
In reality, Africa could leapfrog, not by digitizing paper, but by building learning systems. But leapfrogging is not automatic. If we adopt AI without owning the compute, data, and regulatory backbone behind it, we will fall deeper into dependency than ever before. We will rent intelligence instead of producing it. And that is how the trap becomes permanent.
And We Cannot Wait for Policy to Mature. We Must Lead Policy There.
Fintech did not scale because Kenya redesigned regulation. Regulation evolved because fintech became too useful to ignore. The same must happen with AI in health. Startups must build systems government can inherit, not pilots it can admire. And platforms that become standard, rather than optional.
The Thesis is Simple:
We either cross the chasm, or we become experts at living inside it. The next couple of years will decide whether Africa uses AI or authors the future of health intelligence. The window is open. This time, we must not step halfway through. We must go all the way. DPE hopes to lead the way.