Walk into most organisations that have "adopted AI" and you will find a curious split. On paper, the tools are everywhere. In practice, a handful of enthusiasts use them constantly, a larger group dabbles, and a quiet majority has gone back to working the way they always did. The licences are paid for. The behaviour has not changed. This is the adoption gap, and it is one of the most expensive gaps in business right now.
The instinct is to blame the tool, the training, or the people. The more useful place to look is the manager. Adoption is not a distribution problem to be solved by sending a link. It is a change of behaviour, and behaviour change is led, not announced.
The Adoption Gap Is a Leadership Gap
The Microsoft and LinkedIn 2024 Work Trend Index found that most knowledge workers already use AI, and that a striking share bring their own tools to work without telling anyone. Read that carefully. The problem is usually not reluctance to use AI. It is that people use it in the shadows, inconsistently, and without the shared norms that turn individual hacks into team capability. Where a manager leads adoption well, private experimentation becomes a shared practice. Where they do not, it stays hidden and fragile.
Adoption depends on things only a manager can provide: a clear reason this matters for our work, permission to spend time learning, and visible proof that the boss is doing it too. When those are missing, no amount of platform investment closes the gap.
Six Behaviours That Make Adoption Stick
Across the teams we work with, effective adoption leadership tends to show up as a small set of concrete behaviours rather than a grand strategy.
Set the mandate and the narrative. People need to know why, in terms of their own work, not the company's press release. A good mandate connects AI to something the team already cares about: less drudgery, better quality, more time for the work only humans can do.
Model it visibly. A manager who talks about AI but never shows their own use is asking the team to take a risk they will not take themselves. Show your prompts. Show where the tool helped and where it let you down.
Make experimentation safe. Amy Edmondson's research on psychological safety is unambiguous: people only try new things when they believe an honest mistake will not be held against them. Adoption is a stream of small, imperfect attempts. Punish the mess and you kill the learning.
Have the support conversations. Adoption surfaces anxiety, and anxiety that is not named does not disappear, it just goes quiet. The best managers ask directly what is getting in the way and coach rather than mandate.
Set responsible use norms. Freedom to experiment needs guardrails people can actually follow: what data is fine to use, when a human must check the output, and where AI should not be used at all.
Embed and sustain it. Adoption that is not built into how work actually happens fades within a quarter. Sustaining it means revisiting workflows, celebrating real examples, and keeping the conversation alive well past the launch.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
From Reading to Doing
If those six behaviours feel like a lot to hold in your head, that is exactly why we turned them into tools. The AI Adoption Leadership Diagnostic scores you across five dimensions of adoption leadership and returns a personalised action plan. The Leading AI Adoption Reference Card puts the behaviours on a single page you can keep beside you, and the Leading AI Adoption Development Planner turns your results into a ninety day plan. All three sit inside the Adopt module of the Augmented Leader Suite.
Try This
Before your next team meeting, write down the honest answer to one question: in the last two weeks, what have I visibly done with AI that my team saw? If the answer is nothing, that is your adoption gap in miniature.
Then pick one workflow your team runs every week and redesign it together, out loud, with AI in the mix. Decide what the tool does, what a human does, and who checks the result. One shared, visible workflow does more for adoption than ten licences handed out in silence.