Your company bought AI licences. Here's why nobody is using them.

A Ferrari, ridden like a bicycle. That's the picture I can't shake when a company shows me its AI licence usage numbers. The purchase was announced with a town hall. Six months later, sign-ins are flat and the finance team is asking why the line item exists.

Before I trained people on AI, technology adoption was my actual job. I spent a decade inside a global pharma company accountable for rollouts: a CRM into 15+ countries and 300+ users, analytics tooling worldwide, a data literacy programme across the whole business. I've chaired the meeting where a rollout quietly died. Every failure mode I now get hired to fix is one I was once paid to prevent, with a budget and a board watching.

So when adoption flattens after month two, I don't think your people are lazy. I think one of four familiar things happened.

Nobody owns it

The licences were bought by leadership, handed to IT, and championed by no one. Ask a simple question: who, by name, is accountable for usage in March? If the answer is a committee or a department, the answer is nobody. Every rollout I've seen work had one person whose week got worse when usage dropped.

The middle went quiet

Middle managers make or break every tool, and they're the group with the least airtime in AI conversations. They're asked to push a tool they haven't mastered to teams watching them for cues, while delivering the same targets as last quarter. Silence is the rational move. Give them something specific instead: a working use case for their own team meeting, built with them first, so the first time they present the tool they're demonstrating their own time savings rather than reading slides written by someone else.

The pilot proved nothing

A pilot that ends with a slide deck changes no behavior. A pilot that should have been called an experiment ends with a workflow someone still runs every week. If your last pilot produced a readout and no running workflow, it wasn't a pilot. It was theatre.

Training taught the tool, not the task

A two-hour generic introduction produces two days of curiosity. What holds is one hated task per person, rebuilt with the tool, finished inside the session. People don't adopt software. They adopt the version of their Tuesday that got easier.

What I'd do with a flat usage curve

Name the owner. Find the five people already using AI on personal accounts, because they exist, and make them champions instead of secrets: the shadow users are your head start. Get each team one working use case of its own, built hands-on, starting with whatever task the team complains about most. Then track weekly active usage per team and let the number be visible. Sentiment surveys tell you people find AI "interesting." Usage tells you whether the Ferrari is leaving the garage.

None of this is glamorous. Neither was rolling out a CRM to field reps in fifteen countries. It worked anyway, and the same discipline works here.