Training 10,000 people is not enough if the work stays the same

Rwanda has just closed a large-scale training programme. The Digital Ambassadors Program, funded by Korean cooperation to the tune of USD 4.5 million between 2022 and 2026, held its final steering committee on 20 May 2026: 1,567,034 citizens trained in basic digital skills, over 93% of the target reached, a thousand ambassadors deployed across communities.

There is a second figure in that report, and it is the more interesting one: among the citizens trained on the public service platforms, use of those services rose from 30% to 73%.

That second figure is rare. The training reports I read almost all stop at the first.

I have helped design learning paths for more than 10,000 people. What I take from it fits in one sentence: training produces a capability, and nothing more. Value appears later, when that capability meets a real task inside an organisation that allows it.

A citizen trained on an online service decides alone whether to use it. An employee depends on three authorisations, none of them theirs: access rights, granted by the employer; the method, approved by the manager; acceptance that the work is done differently, which belongs to the organisation. It is that second case the rest of this piece is about.

The figure you publish and the one you never go and get

At the end of a session, three indicators are available the same day: attendance, satisfaction, sometimes a knowledge check. All of them stop at the classroom door. What happens at the desk requires going back three months later, and that is the only place where training becomes something other than an expense.

The Kirkpatrick framework has set out four levels of evaluation since the 1950s: participant reaction, learning, behaviour at work, results for the organisation. In the programmes I see go by, the first two are measured almost everywhere, the last two almost nowhere. Every training manager knows these levels. The problem lies elsewhere: nobody decided, before the training, who would go and look.

Why an acquired skill ends up serving no purpose

The first reason is the calendar, and the gap is striking. What I see most often: three days to a week of training, then six months to a year before the practice actually enters daily work. The cause is nearly always the same — the training was ahead of where their department and their organisation stood.

The second is the one I meet most often at the moment. Someone trains on AI to produce their meeting notes, their reporting, their activity decks. They come back equipped. Their organisation has neither the tools, nor the authorisations, nor a position on its own data. So they take the simplest route: they pay for the subscription themselves and do at home what they cannot do at the office.

This is not disobedience, it is the reflex of a motivated employee working around a process that takes too long. But in a bank, a hospital or a sensitive public body, it means organisational data flows through a tool nobody approved, under a personal account, with no contract and no traceability. The training worked perfectly. The risk comes from everything that was not prepared around it, starting with a clear position on data: where it goes, who holds it, who the organisation is willing to entrust it to. That call does not belong to the learner.

The third is the simplest: nobody decided what should change. Training happens because a budget exists, because a partner offers it, sometimes because a senior manager heard about it at a conference. Each of those reasons explains the spending without describing a single expected outcome. If nobody can name, before the session, the specific task that will be done differently, the answer is already known.

These three causes share one trait: they all sit in the work environment waiting for the learner on their return, never in the quality of the training.

Change is collective, or it does not happen

An organisation that wants to absorb an innovation — AI today, a management tool yesterday — cannot place it on the shoulders of the few people it sends to training. When one department moves and the others stay put, the gap widens. I have seen those gaps produce organisational and HR problems that cost far more than the original training.

Two responsibilities answer each other here, and neither cancels the other.

The manager's, first. A manager's role is to facilitate, to drive the strategy, to look after the effectiveness and the wellbeing of the group. Being the most knowledgeable person on the team is no part of it. Yet this is where I see the most stubborn blocker: a manager who feels they must demonstrate that they know, and who therefore cannot welcome a skill someone else brought back. That ego costs more than the training did.

The learner's next, and it is widely underestimated. Someone who comes back trained holds new knowledge, and part of the work of spreading it falls to them: explaining what they learned to the people it concerns, including when that is unwelcome. But talking about adopting a tool means talking about budget, security and process — three subjects they neither have in mind nor see as a whole. They are asked to carry a change without being given the vantage point that would let them carry it.

When that spreading does not happen, I hear almost always the same reasons: not feeling sufficiently involved in the organisation, or being convinced that their voice will not be heard — past failures, sometimes indirect censorship.

Four decisions to make before the first session

Competency-based design starts from real activity rather than from a syllabus. Taken seriously, it forces four decisions before anything is designed.

1. Which task will be done differently. Not "bring AI into the team", but "meeting notes will be produced and reviewed by the person facilitating, using a tool the organisation has approved". Framed that way, the task becomes verifiable, and the programme to build becomes obvious.

2. Who has to authorise the change. Almost never the learner: their direct manager, and often beyond — security, procurement, leadership. If there is one warning sign to watch, it is that one.

3. When the first real opportunity to practise will happen. A date, not an intention. Ideally within two weeks, on a task that matters, with a consequence if it goes wrong.

4. What will be looked at three months later, and who will look. One indicator is enough, provided it measures the work: how many sets of meeting notes were produced the new way, with the approved tool, over the last quarter. A number, verifiable, with no survey involved.

Where I have had these four decisions taken, they rarely required more than an hour, and they weighed more on the outcome than any additional module.

A fifth practice, cheap and disproportionately useful. At the end of a programme I produce two documents rather than one: a summary note for the learners, and — in large organisations — a short note, sometimes a few-minute presentation, for the leadership. What was learned, what it makes possible, what it now requires. Learners leave equipped; managers know what they now have to authorise. It is the minimum for knowledge to circulate instead of staying stuck on the floor where it landed.

Professional scale changes the difficulty

In February 2026, the Ministry of ICT and Innovation signed a memorandum of understanding with Oracle to open access to certified training in cloud computing, artificial intelligence and database technologies. This leaves citizen literacy for professional skills, and the three authorisations from the start of this piece become the central question. In the organisations I work with in Kigali, certified training is often bought before anyone has checked that the workplace will follow.

The takeaway

A national programme went and measured what its learners actually did after the training. That is a standard few organisations hold themselves to at their own scale, where it is far easier to obtain.

A training programme is judged by what is done differently at the desk three months later. That outcome is decided before the first session: the task that must change, the people who must authorise it, the date of the first real use, the indicator someone will go and check.

Coming back with a certificate transforms nothing on its own. What turns an individual skill into a collective capability is how it is supported: who is told, who authorises, who checks.

---

If you are preparing a training programme and want to make sure the work will actually change, let's take 30 minutes. No pitch — just a conversation.