Most organizations measure their AI progress by adoption: how many staff are using ChatGPT, how many departments have a pilot running, and how many workflows have been “AI-enabled.” AI adoption is visible, easy to report in a board meeting, and almost beside the point.
The harder question is ownership. Not “are we using AI” but “whose AI are we using, on whose terms, and what does that cost us when it matters most?”
Three Conversations, Only One of Them Builds Anything
In our earlier piece on Africa’s AI future, we laid out something worth repeating here because it changes how you should read every AI headline: the world isn’t having one conversation about AI; it’s having three, and they move at different speeds.
The first is about responsibility—ethics, safeguards, and inclusion. Important, but it’s a conversation about how to behave well as a user of someone else’s system.
The second is about infrastructure and market power: compute, data pipelines, and model training capacity. This is the conversation playing out in Washington, Beijing, and increasingly in Nairobi’s new data centers and Rwanda’s AI governance ambitions.
The third is about sovereignty and productivity: who owns the datasets, who controls the deployment, and who builds the platforms. This is where Masakhane’s African-language datasets and South Africa’s Centre for AI Research are doing the real work.
Most African organizations—KCL included until we deliberately corrected for it—default to the first conversation almost by habit. It’s the safest one to have. It’s also the one that produces compliance frameworks instead of capability.
What “Renting” AI Actually Costs You
Here’s the part that doesn’t show up in a pilot-program slide deck.
A foreign-trained model doesn’t know that “informal sector” describes most of Uganda’s economic activity, not an edge case. It doesn’t know what a loan officer in Mbarara actually needs to assess creditworthiness for a trader without a credit history. It wasn’t built for an agricultural extension worker advising in Luganda or Swahili. Every time an organization deploys a general-purpose model against a hyper-local problem, someone quietly absorbs the gap between what the model assumes and what the market actually looks like, usually the end user and sometimes the institution’s own credibility.
There’s a second, quieter cost: leverage. An organization that only consumes AI has no seat at the table when pricing changes, when a vendor’s terms shift, or when a government needs to negotiate data residency requirements. Capability is what turns a user into a negotiator.
This Isn’t Theoretical; It’s Already the work.
KCL’s Head of AI, Dr. Teddy Nalubega, sat on the “Responsible AI in Public Service Delivery” panel at the OECD–African Union AI Dialogue 2.0 in Cairo in November 2024—a gathering of more than 30 countries working through how to actually implement the Continental AI Strategy, not just endorse it. Her position there, and the one we carry into every client engagement, isn’t to deploy faster; it’s to deploy it right, with mechanisms that guarantee fairness, accountability, and the kind of long-term public trust that survives a change in government or a change in vendor.
She carried the same argument to the Paris AI Action Summit in 2025: Africa’s AI future has to be written in its own languages, values, and vision, not translated after the fact from someone else’s.
From Compliant User to Capable Owner
None of this requires an organization to build its own foundation model. It requires three honest decisions, made deliberately rather than by default:
- Audit what you actually own. Most organizations don’t know whether their AI vendor contracts give them rights to the data they’re feeding in. Find out before you scale, not after.
- Localize before you automate. A grant-writing workflow, a citizen service chatbot, and a credit-scoring tool—each one needs to be tested against local context before it’s trusted with real decisions, not assumed to generalize because it worked in a different market.
- Build governance alongside capability, not after it. The organizations that will negotiate well in five years are the ones writing their AI policy now, while the stakes are still low enough to get it wrong safely.
That’s the work behind KCL’s AI policy and strategy consulting—not a generic AI readiness checklist, but the same questions of capability, ownership, and governance that are already shaping the continental conversation, applied to your organization’s specific bottlenecks.
If your organization is further along on adoption than it is on ownership, that gap is worth a conversation. Reach out to us at consult@kcl.co.ug.
Discover more from Knowledge Consulting Ltd
Subscribe to get the latest posts sent to your email.
