How to Lead an AI Rollout People Actually Adopt
The pattern is consistent enough now that I can usually predict the second half of the conversation.
The licences were bought eight or nine months ago. There was a launch, a set of training sessions, an internal champion or two who were genuinely enthusiastic. Usage climbed for about three weeks.
Then it settled at a number nobody wants to put in a board pack. A small group uses the tools constantly. Everybody else opened them once, used them to rewrite an email, and went back to working the way they worked before.
The business case assumed adoption. Adoption is the part nobody planned.
The Old Way Still Works
Most AI rollouts are designed as a distribution problem. Procure the tools, provision the accounts, deliver the training, measure the licences issued. That approach works for a new expenses system, where the old route is switched off and people have no alternative.
With AI there is always an alternative. The old way of writing the report still works. It is slower, but it is known, and it carries no risk of producing something wrong in front of a client.
The research points at the same gap. Accenture found that 81 percent of employees believe their leaders understand AI, while only 20 percent feel like co-creators of the change. Gallup found that 44 percent of leaders use AI frequently, against 23 percent of the people who work for them. The direction of both numbers is the same. The change was decided and understood at the top, and the people expected to carry it out are standing outside it.
Patrick (name changed, details adjusted) was a Country Manager at a financial services firm and had championed the rollout personally. He used the tools daily. He assumed the flat usage numbers meant the training had been poor, so he commissioned more training.
Usage moved for two weeks and then returned to exactly where it had been.
His people knew perfectly well how to open the tool. What they had never been given was a reason to take the risk of using it on anything that mattered.
The Fear Nobody Puts in the Survey
Ask an engagement survey whether people are concerned about AI and you will get a manageable answer. Some concerns about data protection. Some questions about which tasks are appropriate. Nothing alarming.
Sit with the same people in a workshop with their manager out of the room, and a different conversation happens.
What people are actually working out is whether this tool is a better way to do their job or the first visible step towards not having one. Almost nobody says that in those words, to anyone, ever. It surfaces instead as scepticism about accuracy, sudden rigour about compliance, and a stated preference for doing it properly by hand.
I want to be careful here, because this gets misread as irrationality that needs to be managed away. Someone who has spent fifteen years becoming the person who writes the best analysis in the department is being asked to adopt a tool that does a passable version of that in nine seconds. Their hesitation is an accurate reading of their own situation, and telling them the change is exciting does not address it.
The organisations where adoption moves are the ones where a senior person says the difficult part out loud. That the work will change. That some tasks will go. That the intention is to move people up rather than out, and here is what that specifically means for this team. People can work with a hard answer. Enthusiasm that skips the question leaves them exactly where they were.
What Actually Moves Usage
Four things move usage, in the order they tend to matter.
The first is a reason that survives being repeated. A short, plain explanation of why the organisation is doing this, which a team lead can say in their own words to eight people without a slide. If your leadership team produces eight different versions of it, that is the first thing to fix, and it is a half-day of work rather than a programme.
The second is permission to be bad at it. Adoption is a competence risk before it is anything else. Senior people avoid new tools in front of their teams for the same reason they avoid speaking a foreign language badly in a meeting. The fastest intervention I know is a visible senior person using the tool imperfectly in public and saying so out loud. It costs nothing and it changes what the room believes is safe.
The third is the middle managers, and specifically their numbers. If adoption costs a manager throughput this quarter and pays back next year, they will do it next year. The incentive is working exactly as it was designed to work, and no amount of communication changes an incentive.
The fourth is scope. Rollouts that succeed usually replace one specific recurring task for one specific group, completely, and let that spread from there. The ones that stall hand over a general capability and leave every individual to work out where it fits, which most people will not do while busy.
The uncomfortable part is that none of this is technical, and none of it can be delegated to the programme office. It is leadership attention, applied to a small number of specific conversations, at a point in the year when leadership attention is usually somewhere else.
That is also why a stalled rollout is recoverable. The tools already work. The people already have access. What is missing has never been the technology.
Sources: Accenture, Pulse of Change (accenture.com/us-en/insights/pulse-of-change). Gallup, Frequent Use