AI adoption is not capability

AI access is easy to measure. Knowing what people are actually capable of doing with it is harder. Andy Lambert on a conversation with Erica Farmer, co-author of AI for People Professionals.

AI capability inside a business can be wildly uneven. It doesn’t map neatly onto seniority, function, or how much has been spent on tools. Some people do genuinely sophisticated work, usually because they taught themselves. A corridor away, whole teams have a licence they’ve opened twice.

The instinct is to treat this as an engagement problem and run more training at it. Unevenness is what you should expect when access is rolled out at scale and nobody is asked to demonstrate what they can actually do. That was the thread of a recent LinkedIn Live conversation with Erica Farmer.

AI adoption often starts with personal productivity

In Erica’s experience, AI adoption is still largely at the personal-productivity stage. Individuals have found what works for them. The best performers get faster, and build small systems for themselves.

The result is a large gap between the people changing how they work and everyone else. Personal gains don’t automatically add up to better performance across the business. Until there’s a shared baseline of capability, the efficiency story stays stuck in anecdotes and licence counts.

And efficiency isn’t the real ceiling. The bigger opportunity is to rethink processes and do work that wasn’t possible before. In Erica’s view, few businesses are operating at that level yet. As she put it: “I’d rather have 20% of the organisation ideating, innovating, using AI as a thinking partner, than 100% rewriting emails with Copilot.”

Bridging the gap is a people problem

Another tool rollout won’t close that gap. The harder task is turning individual gains into changes in how the business works. HR and L&D have an important role here: raising the baseline of AI capability across the workforce and making sure good practice doesn’t stay in private pockets.

Her shorthand was: “Adoption is not capability.”

Many organisations still measure training completion and tool usage. But if the problem you set out to solve was “give people Copilot,” you may have solved it. That was never the problem worth solving.

What does “good with AI” look like?

One question in the conversation was how to judge success when nobody has defined what good looks like.

AcademyAI uses a six-capability model, informed by the Alan Turing Institute’s AI Skills for Business Competency Framework and Skills England research. It covers literacy, safe and responsible use, framing, specification, application, and evaluation and reflection. The broader point is that capability needs to be something you can assess and improve, person by person and team by team.

What people leaders can do now

AI capability work means leadership explaining what’s changing, what isn’t, and what that means for people. If you don’t communicate clearly, people fill the gaps themselves, often with fear about their jobs. Be explicit about whether roles will be reshaped or displaced.

Pair AI guardrails with clear permission. A policy alone sets rules on data, risk and governance. People also need explicit permission to experiment and a clear sense of what acceptable new work looks like, otherwise they keep doing what they’ve always done, even after training.

Bring more than one function into AI decisions. AI touches different parts of the business at once, and a decision made in isolation in one function can create problems in another.

Use change champions, not only AI enthusiasts. Technical enthusiasm doesn’t necessarily make a good communicator. Change champions need to communicate well, negotiate, think critically and systemically, and have empathy for the change teams are going through. They look more like a critical friend than an AI trainer, and they feed real issues back into governance.

Treat individual AI wins as organisational assets. As Erica put it: “If you’re not sharing so the organisation can benefit from the great work you’re doing individually, that’s as bad as not doing it at all.”

Start with an honest baseline

It starts with an honest picture of workforce capability, rather than training attended or tools used, and a willingness to find out that picture is patchier than hoped.

Watch the full LinkedIn Live conversation with Erica Farmer.

Andy Lambert is co-founder of AcademyAI.