A number that refuses to move is telling you something
One in three workers says they feel properly trained to work with AI. A year earlier, the answer was the same one in three. Through the year AI use climbed hardest, that figure did not shift by a percentage point — across 11,749 respondents in 14 countries, surveyed by the same house on the same instrument.
The first reading of that is a volume problem: more sessions, more coverage, keep pushing. There is something to it. It is also the reading that has not worked. The number sat still through the twelve months in which nearly every large organization at once put training in front of nearly every employee. Whatever went out did not land where it was aimed.
A second survey is specific about what got taught
Research from The Conference Board this July is unusually direct about the content of what most organizations delivered. Sessions concentrated on AI literacy and basic prompting technique. Far fewer helped people integrate AI into a workflow or apply it to the problems their own job actually consists of.
That distinction is the whole finding. A person can finish an AI literacy session, work through the trainer's examples fluently, and leave no closer to knowing what to point the tool at Monday morning. They now understand the tool. The work in front of them is untouched. Asked afterwards whether they feel properly trained, they say no, correctly, having just completed the training.
"Properly trained" is a claim about the path, not the tool
Feeling properly trained is not a measure of what somebody knows about a piece of software. It is a measure of whether they can see the route from the tool to their own work. Generic AI literacy delivered at scale is competent training on the wrong question — the workers surveyed are not asking whether AI matters. They are asking what it changes about how they get through their week.
Which is why the volume argument keeps coming up short. Running the same generic session again produces the same number. The material has to change, and it has to be the team's own work — because that is the only room that teaches the work as opposed to the tool.
A one-question test that costs a leader nothing
The awkward version a leader can perform in fifteen seconds, without any figures at all: has the number in your own head moved? Not the survey number — your own read on whether the training your people went through changed how the work gets done.
If the honest answer is no, that is worth taking at face value. The correct response is not another round of the same material at higher volume. It is a shift in what is inside the room. Your team already has the work; the room is missing it.
We build the material out of your team's own work, and the "properly trained" number moves because the session has something to answer to: Run a workshop on the work your team already has →
Keep reading: How you get good at AI · The rung you learned on
This is the thinking behind a Trusted AI Culture.
Sources: Boston Consulting Group, "AI at Work: Strategy Matters More Than Tools" (Fourth Edition), June 2026, n=11,749 workers across 14 markets. The Conference Board, "Skilling for AI: Critical Factors for Navigating AI Disruption," July 2026. Both are self-reported.
Written with AI in the loop: my idea, AI drafted and sharpened, my judgment on the way out. Every word is mine to stand behind. — Darren