AI Literacy · POV
Compliant but Not Capable.
Why Your AI Training Isn't Producing ROI, and What Does.
If your AI ROI is missing, the likeliest reason is not the models and not the tools. It is that you funded training and expected capability. Training builds literacy: people understand what AI can do. It does not build proficiency: people actually capturing value on real work. Gartner named the gap in one line worth stealing, when leaders fail to make AI adoption social and visible, they risk a workforce that is compliant but not capable. Here is why the training line item does not move the ROI line, and the four moves that do.
30-SECOND POV
- Completion rate is not capability. A 94%-trained workforce can be 0% more productive. Training builds literacy; value lives entirely on the proficiency side, and proficiency is built by doing, not by hearing about it.
- The ROI is missing because you funded the wrong 10%. Most capability comes from practice and peers (70-20), not courses (10). AI programmes invert the spend and wonder why the licences sit idle.
- Four moves build proficiency: mini-challenges instead of workshops, social and visible practice, ring-fenced time, and leading-to-lagging measurement. None is technical. All are organisational, which is exactly why they get skipped.
The completion rate that lies to the board
The most expensive version of this I have watched cost a firm the better part of a million euro across a year, reached 94% of the workforce, scored well on completion and on the quiz, and changed almost nothing. A year later the same people did the same work the same way and the tool licences sat mostly idle. Nobody had failed the training. They passed it. They had simply never once been made to use AI on their own real work under conditions that made the new way easier than the old.
That is the trap in one sentence: the board sees 94% trained and hears 94% capable, and the two numbers are not related. Completion is a compliance artefact. It certifies that someone sat through the material. It certifies nothing about whether their Tuesday looks different. Most AI-readiness reporting in 2026 is a wall of these artefacts, completion rates and quiz scores, presented as capability, which is why so many programmes report success while the ROI the CFO was promised never arrives.
The scale of the underlying gap is not subtle. Gartner\'s research puts roughly 71% of CIOs saying their workforce is not prepared for AI, projects that essentially all IT work will involve AI by 2030, and predicts that by 2027 half of enterprises without a people-centric AI strategy will lose their top AI talent to the ones that have it. The people who are good at this leave organisations that treat readiness as a training catalogue. That is the compounding cost of getting it wrong.
FOUR FAILURE MODES
Where the ROI leaks out
Compliant-but-not-capable is not one mistake; it is the accumulation of four, each individually defensible, that together guarantee the training spend never becomes capability. Run this against your own programme.
The completion-rate mirage
Leadership tracks training completion and reads it as capability. A 94%-trained workforce can be 0% more productive. Completion is a compliance artefact; it says nothing about whether anyone's Tuesday changed.
Funding the wrong 10%
Budget goes to formal courses (the 10% of how adults learn at work) while the 70% on-the-job practice and 20% social learning, where proficiency is actually built, get nothing. The spend is real; the behaviour change is not.
Unprotected learning time
Practice time is whatever is left after the real work, so there is never any left. The one move that reliably separates capable workforces from compliant ones, a manager holding an hour a week, is the one most often skipped.
Watching only lagging metrics
The board asks for cycle-time and revenue-per-head at two quarters, they have not moved because they never move that fast, and the programme is cancelled one quarter early. No leading indicators means no nerve to hold the line.
What to fund instead
The fix is four organisational moves, and the reason they are hard is precisely that none of them is a purchase. Replace the workshop with the mini-challenge: a ten-to-fifteen-minute task on someone\'s real work that returns one useful result, so the barrier ("I do not have time to learn a tool") becomes a spark ("I already got one thing done"). Make practice social and visible through a community of practice and a shared prompt library, so each win pulls the next person in. Ring-fence the time, an hour a week on the calendar, defended, or delivery pressure eats it. And measure the full spectrum, with leading indicators (challenges completed, prompts shared, weekly tool usage) feeding the lagging ones (cycle-time, ticket deflection, output per head) the board actually wants.
The operator framework behind all four, with the metrics ledger, is written up in full at Workforce AI Readiness, and the runnable practice, our own free, non-gated library of 24 mini-challenges by function, is ready to drop into a "challenge of the week" cadence. The distinction to hold onto is the whole game: literacy you can buy, proficiency you have to build. The organisations that will have a genuinely AI-capable workforce in 2027 are not the ones that trained the most people. They are the ones that got the most people to change one workflow, then another, in public, with the time protected and the leading indicators on a wall where everyone could see them.
Compliant but Not Capable: Frequently Asked Questions
What does "compliant but not capable" mean for AI adoption?
Why does AI training spend so often produce no ROI?
What actually builds AI proficiency instead of training?
How should a CTO or CAIO measure AI workforce capability?
Who owns AI workforce readiness in an enterprise?
Build capability, not just compliance
The framework, the metrics ledger, and the runnable challenge library, everything you need to turn training into proficiency.