A seven-level ladder of AI capability from awareness to practice-shaping, and the case for judging training by how many staff reach Level 3, independent competent use.
We wrote before about why generic AI training does not change the work: staff complete the course, and nothing is different a month later, because a course teaches the tool and the tool was never the obstacle. This piece names the actual ladder training is supposed to climb, and the specific rung below which nothing durable has happened yet.
None of what follows is complicated. It is a plain list of seven capability levels, in the order most people actually move through them, and a case for treating one particular level as the minimum a training programme should be judged against.
Fewer levels and the model cannot tell the difference between someone who can follow a written prompt and someone who can judge, unsupervised, whether a tool's output is safe to use. More, and the ladder turns into a certification scheme nobody outside L&D can hold in their head. Seven is the number that separates dependence on another person from genuine, transferable competence, without fragmenting into a checklist nobody remembers past the training session.
Levels 1 and 2 both depend on somebody else being present, physically or in the prompt they wrote. The moment that person leaves the room, the capability leaves with them, and what looked like training success on the day quietly reverts within weeks. Level 3 is the first level that survives without a hand on the tiller: the person can do the task correctly, alone, and knows unprompted when to be suspicious of what came back. Everything below that is a demonstration of the tool, not a capability the organisation actually owns.
A capability that disappears when the trainer leaves the room was never a capability. It was a demonstration.
This is the same mechanism we described in generic training: a course that stops at guided use produces confident completion certificates and no lasting change, because certificates measure attendance, not the point at which the trainer's presence stopped mattering.
Training to Level 3 for every relevant tool costs more time and more follow-up than training to Level 2 and calling it finished, and it produces a slower, less impressive-looking rollout in the short term. The honest answer is that Level 2 training is cheaper precisely because it defers the real cost to later, when the capability turns out not to have survived the trainer's departure and the organisation has to redo the work anyway, this time having already spent the goodwill and the budget once.
Measure training success by how many staff reach Level 3, independent competent use, not by how many completed the course; below Level 3, the capability leaves the building with the trainer.
No external statistic cited; this article defines the organisation's own training maturity model rather than presenting third-party evidence.
If this is the question on your desk, a thirty-minute call tells you whether the service fits, or that you do not need us yet.