Rwanda now has a national AI agency. Who owns AI in your organization?

In June 2026, Rwanda's Cabinet approved the creation of a National Artificial Intelligence Agency — the country's first institution fully dedicated to AI. Its mandate: accelerate AI adoption, investment and governance across every sector.

Beyond the announcement itself, the signal is clear: Rwanda has decided that AI is too important to be left to scattered initiatives. It needed an owner. And a framework.

Now ask the same question at the scale of your own organization: who owns AI where you work?

What is actually happening inside organizations

There is a paradox I keep running into in my training sessions: most people already use AI — in their personal lives. They ask ChatGPT for advice on a relationship, help with a purchase decision, an administrative letter, a legal question. At work, though, few make the jump to a professional approach. Not for lack of tools — for lack of understanding of what actually happens behind ChatGPT, Claude or Gemini.

The result: inside organizations, AI is already in the building — it came in through the side door. A program officer has been drafting reports with ChatGPT for months; their management has no idea. Nobody ever asked, that's all.

Nobody decided on this adoption. It simply happened.

Why this is a real problem

This quiet improvisation gets expensive. Three people on the same team solve the same problem separately, with three different tools — what one of them learns never reaches the others. Data leaves without control: a contract or a budget pasted into a consumer tool whose terms nobody has read. For an NGO handling beneficiary data, that is the kind of detail that surfaces during a donor audit — not before. And the day the person who figured out the good method leaves, their know-how leaves with them.

And there is a quieter cost still: badly used AI becomes agreeable. Feed it vague, unstructured requests with no real thinking behind them — and it will side with you. Being helpful is what it was built for; pushing back has to be asked for. The less you challenge it, the less it challenges you. That loop settles in silently, and it turns a decision-support tool into a machine for validating your own ideas.

The more resourceful your team is, the faster the problem grows.

What a structured organization does differently

Ban it? Too late — AI is already in daily use, and it is a productivity lever you would be wrong to throw away. Launch a six-figure "AI transformation program"? Nobody needs one to get started. Rwanda just demonstrated the useful move, at the scale of a country: name someone, set a framework.

In practice, four decisions are enough.

1. Name a lead. One person, not a committee — and not necessarily a technical profile: what matters is understanding how your teams actually work. Name them this week, announce it in one sentence at your next team meeting, and block two hours a week for the role. A mandate without dedicated time is decoration.

2. Pick two or three priority use cases. No audit required to find them: ask in a meeting who already uses AI, and for what. Thirty minutes is enough — the priority use cases are almost always the ones already running under the radar. Prove the value on a narrow scope before expanding.

3. Write simple rules — one page, maximum. What can go into an AI tool, what never does (personal data, confidential documents), which tools are approved. One page people actually read beats a thirty-page policy everyone ignores.

4. Train teams progressively. One 90-minute session per use case, one every two weeks, built on real working documents. Everyone walks out with something they will use the next morning — not a slide deck. This is also where people learn to structure their requests: the whole difference between an AI that challenges you and an AI that just agrees.

Start small — but start structured

A twenty-person organization can put this framework in place within a month. Week 1: the lead and the one-page rules. Weeks 2 and 3: the first two use cases. Week 4: the first training session. What matters is the sequence — an owner, priorities, rules, training, in that order.

That is exactly the Rwandan government's logic with its new agency: before multiplying projects, structure the governance. What works for a country works for a fifteen-person team.

AI is an accelerator. Someone still has to hold the wheel.

---

Your organization is already using AI — the real question is whether that's by design or just happening to you. If you want an outside look at where you stand, let's book 30 minutes. Thirty minutes, no slides.