There’s a question we keep hearing in one form or another, usually from the top of the organisation rather than the middle of it: we rolled out the AI training, so why hasn’t anything actually changed? The training was well attended or the training was mandatory and the feedback was good. The budget was signed off without much of a fight, because everyone agreed AI was too important to sit out. And yet, here we are months later, when anyone actually looks at how people are working day to day, it looks almost identical to before and leaders keep asking the same questions, “why hasn’t this worked” or “what’s next?”
We’ve spent all summer pondering this problem, talking to the faculty members and building solutions to what we now call “the stall phase”, and it is remarkably common. So common in fact, that if you’re living in it right now, the first thing worth knowing is that you are not alone.
It’s worth being precise about what the stall phase isn’t. It isn’t a sign the training was bad or badly delivered, or that people weren’t paying attention, or that the L&D team dropped the ball somewhere. By every measure training is actually built to hit: attendance, completion, satisfaction, it usually did exactly what it was asked to do, and everyone got their trophy. The stall phase shows up regardless of all this, which is the first clue that the thing being asked of it was never really achievable through training alone.
Somewhere in the last decade, most of us relearned a lesson youth sport had already been arguing about long before AI turned up: if you reward people for participating rather than for what they actually achieve, you don’t get more achievement, you just get more participation. Everyone gets a trophy, nobody gets noticeably better, and eventually “taking part” and “improved” stop meaning anything different. AI training has quietly built itself into the same trap, just with better production values and a real budget line behind it. Attendance is an easy metric, so it becomes the thing that gets counted and celebrated, right up until someone senior asks why none of it shows up in the numbers.
And the numbers say it all, this stall phase is the norm, not the exception.
Kyndryl’s 2026 People Readiness Report, surveying 1,100 senior business and technology leaders across eight countries, found 57% of enterprises now have AI properly embedded in their core business processes, up sharply from 35% the year before. But only 11% of those same organisations had hit both of their top two AI-related goals, and only 32% had managed even one. Deployment nearly doubled in a year. Outcomes didn’t.
Stanford Institute for Human-Centered Artificial Intelligence (HAI)’s 2026 AI Index tells almost the same story from a different angle.
Organisational adoption of AI sits at 88%, and 70% of organisations are using generative AI in at least one function. And yet, by that same report’s figures, fewer than 10% of organisations have actually managed to fully scale AI in even a single function.
The training clearly happened, at real scale, to get adoption that high. It did exactly what training was always going to do, and then it stopped, because it hit the edge of what training was ever capable of doing.
Here’s the part worth sitting with: this was never really an AI problem. Treating AI as a tool and talking about AI in a workshop doesn’t mean anything else in the organisation gets changed. The accountability structures are the same as they’ve always been. The people modelling old habits are still modelling them. The workflows nobody’s redesigned are still the workflows everyone’s working around. Wondering whether AI is “working” in an organisation is really wondering whether change management is working in that organisation. Nobody’s reinvented the wheel here, it’s just been refitted to a newer shinier vehicle that still needs someone steering it.
Now, let’s begin the real work, what actually closes the AI adoption gap? The answer is slower and less glamorous than another training session, which is exactly why training gets bought first. Redesigning organisation from the core with AI in mind, real accountability for using the new ways of working, leaders visibly changing their own habits before asking anyone else to, workflows genuinely redesigned so the old path is harder than the new one, and enough hands-on support that people get through the awkward “worse before it gets better” phase. None of that is an AI capability, it’s a change management capability, and it’s usually the thing that was never funded, while the budget went into another round of workshops instead.
This is the exact stall point The Gen AI Academy has been building towards over the past 6 months, and our ambition is a slightly unusual one for a business to admit to: we want to work ourselves out of a job. We measure our own success by whether the organisations we work with need us less over time, not more, which is why so much of what we build is designed to create peer-to-peer learning inside the organisation itself, so the good habits keep compounding long after we’ve left the room.
If the question in your organisation right now is “why hasn’t the training translated,” the honest answer usually isn’t about the training at all. It’s about everything that was always going to need to change alongside it.
That’s enough theory, this is how we’ve built for true AI adoption.
Our AI Adoption Programme is built around the sequence most training skips entirely, going back to basics and diagnosing where an organisation actually is before anything gets deployed, embedding the new ways of working once it has, and measuring whether behaviour has genuinely shifted rather than whether a course got finished.
For the people at the very top, our CEO Coaching works to align the operating model with the CEO and our Leadership Coaching applies the thinking because nobody downstream changes how they work while watching leadership carry on exactly as before.
For organisations that want that capability built from the inside out rather than borrowed indefinitely, our AI Pioneer Programme takes a cross-section of people through a structured, multi-expert journey from AI novice to certified AI Pioneer, so there’s a genuine bench of internal AI champions doing the peer-to-peer teaching we mentioned above, long after any external programme has finished. It’s a qualification, earned over months rather than an afternoon, which is rather the point of everything we’ve just said. Nobody graduates from it for having attended.
If you were to ask me is there one clear pathway for organisations, then the answer is simply no, fundamentally every organisation is built differently with their own leadership structure, org models, employee appetite for change, and budgets. Therefore each journey is unique, and there is no one size fits all.
This AI adoption path is not training in the sense the stall phase usually defines it. This path asks for time, and it asks for effort, in the places that don’t show up on dashboards, which is exactly why so many organisations skip past it in search of something faster. But the reward was never meant to arrive at the start. It comes after the harder work, to the people who actually did it, not to everyone who simply turned up. That’s the real medal worth earning, not the participation trophy.
Sources:
https://www.kyndryl.com/gb/en/insights/people-readiness-report-2026
