Why doesn’t my team use the AI tools we pay for?

Buying the seats is the easy part. I hear a version of the same thing from owners all the time: we bought the licences, two or three people use them every day, and everyone else opened the tool once and went back to doing it the old way. The owner reads that as resistance. When you look at what workers actually say, it’s mostly something more fixable than that.

What do employees say when you ask why they don’t use AI?

Pew Research Center asked 5,273 U.S. workers in October 2024. Among those who don’t use AI chatbots, 36% said a major reason was that there’s no use for them in their job, and 22% said they weren’t interested. Only 10% said they didn’t know how, and 9% said their employer didn’t allow it.

Read that first number carefully. "There’s no use for it in my job" often means nobody showed them the use in their job. A person who spends an hour a day writing the same follow-up emails doesn’t look at an empty chat box and think "that’s my follow-up emails." They see a blank screen and a vague sense that it’s for somebody more technical. The tool is general. Their work is specific. Somebody has to connect the two, and in a small business that somebody is usually you.

Is my team waiting for permission?

Many are. In the same Pew survey, half of workers who aren’t self-employed said their employer neither encourages nor discourages AI chatbots. 12% said their employer encourages it and 8% said it discourages it. Silence from the owner reads as risk to an employee, and most people won’t gamble with client data on a guess.

Think about it from their side. Paste a client’s file into the wrong tool and you might be the person who caused a problem. Don’t use it and nothing happens to you. Until someone with authority says what’s allowed, what isn’t, and which account to use, doing nothing is the safe move. It’s a reasonable reading of the incentives you set.

Did we buy the tool and skip the rollout?

Most businesses did. In the Census Bureau’s AI supplement, fielded November 2025 to January 2026, 64% of firms using AI reported no organizational adjustments at all. Training staff and developing new workflows were the most common changes, and each applied to only about 15% of AI-using firms. The software arrived and the work around it stayed the same.

This isn’t new. The Census Bureau’s first look at the question, covering late 2023 and early 2024, found 50.5% of AI-using firms had made no changes to use it, and 20.8% had trained existing staff. Two years later the tools are far better and the rollout habits look the same. Layering tools on top of an unchanged process doesn’t create different results.

Are they afraid AI will replace them?

Some are, and saying nothing makes it worse. Pew found 52% of workers were worried about how AI will be used in the workplace, and 32% expected it to mean fewer job opportunities for them. Census data points the other way so far: only 2% of firms reported AI-related employment decreases, and 66% of users use AI only to augment tasks.

People don’t get enthusiastic about a tool they think is auditioning for their job. Say the plan out loud: nobody gets replaced, everybody moves up. Then make it true in what you hand over first. If the first thing AI touches is the part of someone’s week they hate, they’ll defend it. If the first thing it touches is the part they’re proud of, you’ll get the silence back.

What gets a team using AI within a month?

Assignment, permission and a short feedback loop. Give each person one named task from their own week, write the rules on one page, use it yourself in front of them, and spend fifteen minutes a week on what worked. None of that needs a training budget. It needs the owner to decide what the tool is for.

1. Pick one task per person that they already dislike, and write it down with their name next to it. Follow-up emails, meeting notes, first drafts of proposals, summarizing a long document.
2. Put the rules on one page: which tool and which account, what data never goes in, and who checks the output before it reaches a client.
3. Use it yourself first, on a real task, where the team can see it. Owner behaviour is the permission slip nobody has to ask for.
4. Sit with each person for thirty minutes on their own task, in their inbox, on their document, on an ordinary Tuesday.
5. Hold fifteen minutes a week: one thing that worked, one thing that broke, one new task to try. Write the wins down so the next hire starts from them.

If you want a way to pick those first tasks, the same logic applies to agents as to chat tools: start where the work is repetitive, rule-shaped and nobody’s favourite.

When is low usage the right result?

Sometimes a role really doesn’t have much for a chat tool to do. Someone who spends the day on a job site, in a treatment room or on the phone with a patient might use it twice a week, and that’s fine. Judge usage against the tasks you assigned, not against a dashboard number. The question worth asking is whether the tasks you named got faster or better, not whether everybody logged in.