Almost every organisation I speak to describes the same problem. AI capability inside the business is wildly uneven, and the unevenness does not map neatly onto seniority, function, or how much the company has spent. There are individuals doing genuinely sophisticated work, usually because they taught themselves, and there are whole teams a corridor away who have a licence they have opened twice.
The instinct is to treat this as an engagement problem and to run more training at it. But I'm not sure that's correct. The more uncomfortable explanation is that unevenness is exactly what you would expect when access to AI tools is rolled out, and nobody is ever asked to demonstrate what they can do.
That was the thread Erica Farmer and I pulled on in our LinkedIn Live. Erica has spent years in HR and L&D, has just released AI for People Professionals, and is one of the most well-respected UK voices on what people leaders can do to help embed AI in day-to-day work.
Where are we today?
If you look at where we are on the trajectory, most organisations are still in phase one: personal productivity. Individuals have found the things that work for them. The best performers get faster. They amplify good practice. They build little systems for themselves. That is not a bad place to start. It is also how you create a large delta inside the organisation between the people who are transforming their work and everyone else.
Personal gains do not automatically add up to organisation-wide efficiency. That is the floor we keep talking about and rarely reach: shared ways of working, best practice that spreads, a baseline of capability across the workforce rather than a handful of A players carrying the change. Until that floor exists, the efficiency story stays stuck in anecdotes and licence usage counts.
And we should be more honest with ourselves about the ceiling. Efficiency and productivity are not the real goal. The real goal is organisation-wide transformation: how we completely rethink a process, what new things become true once new capabilities are unlocked. Most businesses are not there yet, or they are exploring it only in a lab environment. As Erica said: "The real goal isn't efficiency at all. The real goal is organisation-wide transformation. We're still on the starting line on the track."
So the pragmatic move is not another tool rollout. It is to focus on the blocking problem. The reason sits in that gap between individual productivity and organisation-wide change. Bridging it isn't squarely an HR and L&D job, but these roles have just become even more crucial. The real work is now: uplift the baseline of organisational AI capability, build fundamental understanding, and do the unglamorous work of skill-sharing so wins stop living in private pockets.
Erica made the point plainly: "Tool access does not make people AI literate. Will doesn't equate to skill, and adoption is not capability."
Most organisations are still measuring training completion and tool usage. I don't blame anyone for this. It makes sense when there's no better objective measure. But quite simply, the objective was never to get everyone using AI. If the problem you set out to solve was "give people Copilot," congratulations, you solved it. But that was never the problem worth solving.
Erica's line on this is the one I keep coming back to: "I'd rather have 20% of the organisation ideating, innovating, using AI as a thinking partner, than 100% rewriting emails with Copilot". A high rollout figure means little if the work stays shallow.
What does being good with AI look like?
We heard this narrative on the LinkedIn Live: "How can we determine success when we don't know what good looks like?"
This is why we keep coming back to the six measurements that we focus on at AcademyAI: literacy, safe and responsible use, framing, specification, application, and evaluation and reflection.
We have written about them before, so I will not labour them here. The point is simply that capability has to become something you can see, compare and improve, person by person and team by team. Usage charts cannot do that job.
None of this is an IT project. As Erica said, it is very different to a systems rollout. Leadership has to get closer: not just their own AI literacy, but sharing what worked, what failed, and setting the tone of psychological safety. You cannot overcommunicate your AI transformation plans with your team, because if you don't communicate, people will fill in the gaps. Usually with fear about their jobs. Address that fear directly and be explicit about whether roles will be reshaped or displaced. Ultimately, it's our job as leaders to win hearts and minds first and give them a reason to care.
So what can businesses do right now?
Erica's IKEA example was perfect: When the Billy chatbot took first-line inquiries, the company did not simply cut the customer service agents. It upskilled them into design consultants working with customers on design advice. Efficiency and a different kind of customer value, not just a headcount cut. Be an IKEA, was her line: "reshape with people, do not pretend the fear is imaginary."
Pair the AI policy with an AI manifesto. The policy sets guardrails on data, risk, and governance. The manifesto gives people permission to experiment and clarifies what acceptable new work looks like. Without that, people will do what they have always done, even after the training, because it does not feel safe to change.
That psychological safety is driven by leadership, along with a cross-functional governance team that includes operations, contact centres, sales, marketing, HR and L&D, IT, legal, and information security. The point is that multiple people need a seat at the table here. Change cannot be enacted in a silo, but the team needs to be nimble enough and have the latitude to drive change and feel psychologically safe in doing so.
One simple action that needs to be endorsed at the highest level is that the organisation needs to deliberately make time for experimentation.
Erica's guidance was to help drive this behaviour change with 'Change Champions'. The crucial distinction here is that these are not the early enthusiasts or the more technically literate. Technical enthusiasm does not necessarily make a good communicator. A change champion is someone that is trained on communication and negotiation, understands critical and systems thinking, and can empathise with the change that teams are being asked to go through. This is a deeply human, cultural challenge, not a technological one, so those change champions need to look more like a critical friend than an AI trainer. You need trusted influencers who can surface real issues and feed them back into governance.
Then treat individual AI wins as organisational assets. Erica was clear on this: "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."
This all begins with an honest baseline of your workforce's capability, not what training they have attended or the tools they have used, along with the willingness to find out that the picture is patchier than you might have hoped. Then uplift the baseline, share the wins, and measure capability uplift.
And that's what we're doing at AcademyAI. You can see what we're up to here: www.academyai.co
We'd love to hear your thoughts, and if you missed the live, you can watch the replay here: https://youtu.be/wGcIvSAKhu0
Thanks, as always, Andy
