AI adoption is a behaviour problem

A familiar pattern plays out in small firms. The owner reads enough to be convinced, buys everyone a licence for a capable AI tool, sends a cheerful message encouraging the team to use it, and waits for the productivity to arrive. A month later, usage reports tell the real story: two or three enthusiasts are using it daily, a handful have tried it once, and most have not opened it at all. The tool works perfectly. Almost nobody is using it. The owner concludes the technology was oversold, when the technology was never the problem.

The problem is behaviour. Conor Grennan, who runs AI adoption programmes and writes about why they succeed or fail, puts it plainly: "the barriers to adoption are not technological but behavioral." People already have the tool. What they do not have is a new habit, and habits do not arrive because a licence did. They are built, slowly, on purpose. Understanding that is the difference between a firm that wastes its AI budget and one that gets something back from it.

The licence-and-pray failure

Buying licences feels like adoption. It is a decision, it costs money, it can be announced, and it produces a satisfying sense of having acted. But access is not use, and use is not habit. Handing someone a powerful tool changes nothing about the dozens of small moments in their day where they could reach for it and, out of habit, reach for the old way instead.

This is why "we gave everyone ChatGPT" so rarely produces the change owners expect. The licence removes the cost barrier and the access barrier, which were never the real barriers. The real barrier is that people have established ways of doing their work, those ways are automatic, and a new tool sitting in a browser tab does nothing to interrupt them. Hoping people will change because the option now exists is the licence-and-pray approach, and it fails almost every time.

Why behaviour is the bottleneck

Grennan's deeper point is that getting value from AI is less about the software and more about a shift in how people approach their work. In his framing, adoption is about changing how professionals think, work, and solve problems, which is a much bigger ask than learning where a button is. You are asking someone to pause an ingrained routine, consider whether a new tool might do part of it better, and risk the discomfort of being slower at first while they learn. People do not do that spontaneously, no matter how good the tool is.

That is good news for a small firm, oddly, because behaviour is something a small team can actually influence. You do not have a thousand people and a change-management department. You have a handful of people whose habits you can see, shape and reinforce directly. The bottleneck is real, but in a small firm it is also close enough to do something about.

Habit cues that work in small teams

Habits form around cues: a regular trigger that prompts the new behaviour until it becomes automatic. The practical work of adoption is attaching AI to cues that already exist in your team's week. Three kinds work well.

Tie it to a recurring task. Pick a job your team already does on a schedule, the weekly report, the Monday planning, the post-call write-up, and make using AI the agreed way to do that one thing. A fixed task is a reliable cue, and one task done the new way every week beats vague encouragement to use AI generally.

Make it visible. Habits spread in a small team when people can see each other doing the thing. Share the good prompt that worked, show the before-and-after, talk about it in the normal rhythm of the week. Visibility turns a private experiment into a team norm, which is where individual habit becomes shared practice.

Lower the friction. Every extra step between the cue and the behaviour kills the habit. Put the tool one click away, have the useful prompts written down and ready, remove the small frustrations that make people give up. The easier the new way is to start, the more likely it survives contact with a busy day.

Access removes the excuse not to start. A cue is what actually makes someone start.

Modelling it from the top

In a small firm the owner's behaviour is the loudest signal in the building. If you encourage the team to use AI but never visibly use it yourself, everyone reads the real message: this is optional, and not for people like us. If you use it openly, talk about where it helped and where it did not, and bring it into how you do your own work, you make the new behaviour normal and safe to try.

This is not about being the most skilled user. It is about being visible. A founder who shows their working, including the times the tool got it wrong, gives the team permission to experiment and a model to copy. Modelling the habit yourself is the single most effective move an owner has, and it costs nothing but the willingness to do it in the open.

The honest limits

Behaviour change is slow, and you should plan for weeks, not days. The enthusiasts will move fast and the rest will move at the pace of habit, which is unhurried by nature. If you expect a transformed team a fortnight after handing out licences, you will declare failure right when the real change is starting to take. Set the expectation that this takes a quarter, not a week, and judge it on that timescale.

There is also a limit to how much you should push. The goal is people reaching for AI where it genuinely helps, not using it everywhere to satisfy a target. Some tasks are better done the old way, and a team that feels forced will perform compliance rather than build a real habit. Encourage, model, lower the friction, and let the genuine wins do the persuading. Pressure produces usage reports. It does not produce habits.

What to do this week

Pick one recurring task and make AI the agreed way your team does it, starting this week. Write down the prompt that works so nobody has to reinvent it. Use it visibly yourself, and talk about it in your normal team rhythm. Then leave it alone for a month and let the habit form before you add a second. Adoption is built one cue at a time, and one solid habit is worth more than a dozen licences nobody opens.

Building those habits deliberately, rather than hoping they appear, is the core of the AI Advisory for Teams work: turning scattered, occasional AI use into a few reliable habits the whole team shares. If your tools are bought but unused, book a business call.

Sources and further reading

• Conor Grennan, "Lowering Barriers to AI Adoption," OpenAI Forum, 2 October 2024. His central claim that what blocks adoption is behaviour, not technology.
• Conor Grennan, AI Mindset newsletter, 2025. More on why adoption is really about changing how people work, rather than learning a tool.