Why your team won't use the AI you gave them

You bought the licences. Three months in, the usage panel says the thing nobody puts on the slide.

Carlos Andrés Ramírez · 2026-08-22

You bought the licences, ran the launch town hall, and promised this would free up time. Three months in, the usage panel tells the truth: twelve of the hundred activated seats get opened, and it is almost always the same twelve people.

The standard response is training. Another webinar, a prompt guide, a Slack channel full of use cases, maybe a leaderboard of who used it most this month. All of it costs budget, bumps the number for two weeks, and then the number drifts back down. Because the problem was never that people didn't know how to use the tool.

The problem is that using it well doesn't pay off for anyone except the company. And no webinar fixes that account.

The symptom

Why won't my team actually use the AI tools I gave them?

You don't get there by explaining the tool better. You get there by changing what happens after someone uses it well, which in most teams today is nothing good at all. Real adoption means someone notices a measurable difference in how that person works, and right now that difference doesn't translate into anything useful for whoever produced it.

Usage spikes launch week and halves the month after, and nobody notices until the quarterly review.
The heaviest users are junior, and the lightest are senior, because seniors already have a method that works and risking it doesn't pay.
Nobody tracks active use per task, only activated seats, so the number reported upward says nothing about what happens below.
The time the tool frees up gets filled with more of the same work, not with anything different or with fewer hours.
Whoever performs best with AI is also the one most convinced their role could be done with fewer people.

The problem underneath

The incentive conflict nobody puts on the slide.

Ask an analyst to use AI to cut their report time in half and you are asking them, without saying so, to prove that half their week is spare. Nobody signs up for that voluntarily. The rational response isn't refusing in public: it's using the tool at half strength, just enough not to look bad in the adoption survey, just little enough that the time saved never shows.

There is a second friction, talked about less. The learning curve has a dip, and in that dip performance drops before it climbs: the first prompts come out wrong, the first review takes longer than doing it by hand, the first AI-assisted report needs more fixes than the last one did without it. If that dip gets measured the same way as any other performance drop, nobody crosses it. And if the time AI frees up gets refilled automatically with more of the same queue, there is no reason to go faster either: moving quicker just means the inbox empties and fills back up.

You are asking your team to use a tool that could make their own job unnecessary, and expecting adoption to come from enthusiasm.
BECOME

What has to be decided

Four answers no licence dashboard gives you.

Safety: What happens to whoever proves their task now takes half the time. If the implicit answer is that headcount is spare, nobody is going to prove it.
Freed time: Where the saved hour goes: fewer hours, a different task with different value, or straight back into the same queue. Without that answer written down, the team fills it by default with more of the same.
The dip: Who absorbs the lower output of the first weeks while someone learns to use the tool properly, and for how long that dip doesn't count against them.
The measure: Active use per task, not activated seats. The second measures purchase, not adoption, and it's nearly all any dashboard reports.

Take the team using the tool least and ask them, without the word 'adoption' anywhere in the sentence, what would happen if their manager saw their task now taking half the time. If the answer makes anyone uncomfortable, that's why the usage panel is still flat. And it was never about training.

Frequently asked questions

Why won't my team actually use the AI tools I gave them?

Because of what happens after someone uses it well, not because of how it's explained. While using it fast reads as proof that headcount is spare, the team will use it just enough not to look bad in the survey. Adoption rises once it's clear what the person gains, not only the company, from every hour AI hands back.

How do you measure real AI adoption on a team?

With active use per task, tracked week over week inside the team itself, not with activated seats. Activated seats measure a purchase; active use per task measures whether someone actually changed how they work. Most dashboards report the first number because it's the easy one to pull, not because it's the one that matters.

What should happen to the time AI frees up on a team?

Left alone, it almost always gets refilled with more of the same queue, so nobody notices any gain because the day still looks just as full. Unless that hour gets pointed at something different and visible, decided before adoption gets measured, there is no real incentive to work faster with AI.

Why do senior staff use AI less than junior staff?

Because seniors already have a working method, and risking it for an unproven tool costs them more than it costs a junior who has nothing established yet to protect. That gap closes only when using AI well is rewarded rather than read as evidence that the senior's own judgement was replaceable all along.

Let's talk about your adoption

About the author

Carlos Andrés Ramírez — Transformation Director

Specialist in business transformation and reinvention. Director of Specialised Programmes and lecturer in Artificial Intelligence at UPC's Graduate School.