Certified is not the same as changed
Certification counts are easy to manufacture. The real test is what people do differently afterwards.
Every large organisation now has an AI certification number. Thousands of employees certified. A leaderboard by business unit. A slide for the town hall.
I understand why. It is a number that goes up, it is easy to report, and it looks like progress. But I have stopped trusting it, and I think leaders should too.
The number measures the wrong thing
A certification tells you someone passed an assessment. It does not tell you whether they use AI in their work, whether they use it well, or whether anything improved because of it.
It is also easier to pass than ever. Many of these assessments can be completed with the help of the same AI tools they are meant to test. When the exam can be passed by the thing it examines, the pass rate stops meaning much.
So the headline count becomes a vanity metric: real effort goes into producing it, and it tells you almost nothing about capability.
Ask the question after the certificate
The useful question is simple and rarely asked: what did people do differently after they certified?
- Which workflows changed?
- Which tasks now take less time, and is the time saved visible anywhere?
- Which AI use cases moved from a demo to daily work?
- Where did quality get better, and where did it get worse?
If a programme cannot answer these, it has trained people to pass, not to work differently.
Capability is pace-limited
There is a second problem. People being certified are also running projects, managing teams and serving clients. Nobody can keep pace with AI through courses alone, however well designed the learning path. The field moves faster than any curriculum.
That means capability grows through use, in the flow of real work, with time set aside for it. A programme that adds learning hours on top of a full workload and then counts completions is measuring how much people could squeeze in, not how much they changed.
What to measure instead
Three measures tell you more than any certification count:
- Active use in real work, measured a month or more after training.
- Workflows redesigned, not just tools added to old workflows.
- Outcomes moved: cycle time, quality or cost on a process someone owns.
If you want a quick read on where your organisation’s AI effort is actually stuck, the AI Scaling Readiness diagnostic walks through the gaps that keep pilots from reaching production. Few of them are about training.
Certificates are fine. Counting them as transformation is the mistake.