Rolling out GenAI licences is easy. Building a workforce that can actually use them well is not.
Across UK organisations, the same pattern shows up. A handful of people teach themselves sophisticated workflows. A corridor away, teams treat "using AI" as asking Copilot to reword an email. That gap does not map cleanly to seniority, tenure, or how much the company has spent on tokens.
HR and L&D often inherit this as an engagement or training problem. The sharper reading is a measurement problem. When nobody asks people to demonstrate what they can do with AI, uneven capability is the default outcome of a wide tool rollout.
That was the practical thread of AcademyAI's LinkedIn Live with Erica Farmer, co-author of AI for People Professionals and a leading UK voice on embedding AI in people work. The conversation is worth watching in full: Watch the LinkedIn Live with Erica Farmer
Adoption is not capability
Erica put three distinctions on the table that cut through most AI dashboards:
- Tool access does not make people AI literate
- Will does not equate to skill
- Adoption is not capability
Those lines matter because most organisations still score the easy numbers: licences issued, training attended, tokens burned, sometimes even AI leaderboards. Those metrics can rise while the quality of work stays shallow.
Erica's sticky test: she would rather have 20% of an organisation ideating with AI as a thinking partner than 100% rewriting emails with Copilot. High rollout figures mean little if the work never leaves the first rung of personal productivity.
Three stages most organisations get stuck on
A useful way to read organisational AI maturity is as three stages:
- Personal productivity: Individuals find what works for them
- Organisation-wide efficiency: Shared practices that actually move the baseline
- Reorganising work around AI: New processes and work that was not possible before
Most organisations are still in stage one. That creates two predictable problems. AI amplifies A-players, so the gap between top performers and everyone else widens. And the wins of the enthusiastic few rarely become organisation-wide efficiency, let alone true transformation.
As Erica said, "the real goal is not efficiency theatre. It is organisation-wide transformation", and most businesses are still on the starting line.
Why "good with AI" has to become measurable
If promotion, bonuses, or informal status start to track AI use (a trend already visible in the market), organisations need something better than usage charts.
Capability has to become visible: comparable across people and teams, improvable over time. One practical frame maps workforce AI ability across literacy, safe and responsible use, framing, specification, application, and evaluation and reflection. The brand of the framework matters less than the discipline of measuring skills rather than activity.
Without that, leaders keep confusing motion for progress.
What people leaders can do on Monday
Pair policy with a manifesto. The policy covers data, risk, and governance. The manifesto gives permission to experiment and spells out what acceptable new work looks like. Without permission, people revert to old habits after the training day.
Build cross-functional governance that includes operations, contact centres, sales, marketing, HR and L&D, IT, legal, and information security, all with enough psychological safety to move.
Protect time for experimentation at the highest level, on purpose.
Appoint change champions, not only AI enthusiasts. Technical enthusiasm is not the same as communication, negotiation, systems thinking, or empathy for change. Champions should look more like critical friends than tool trainers, and they should feed real issues back into governance.
Treat private AI wins as organisational assets. Erica's line: if you are not sharing so the organisation can benefit from the great work you are doing individually, that is as bad as not doing it at all.
Start with an honest baseline of capability, not of training attended or tools opened. Expect the picture to be patchier than hoped. Then uplift the baseline, share the wins, and measure whether capability actually moved.
This is the work AcademyAI exists to support: assess the AI capability gap, close it by role, and prove it moved. Tool access alone will not get you there.
