Your team has ChatGPT licences. Why is nothing changing?

Dovi Levin Founder, Zmanify · 16 July 2026

Because a licence is not adoption. Most teams that pay for ChatGPT or Copilot use a fraction of what they're paying for, not because the tools are bad, but because nobody has shown them what the tools can do for their work specifically. Generic access plus a launch email changes nothing; behaviour changes when people see their own tasks done faster, live.

How big is the gap?

Bigger than most leadership teams realise. Around 70 per cent of the Fortune 500 have bought Microsoft 365 Copilot licences, yet Recon Analytics' 2026 research found that only around a third of employees with access to a workplace AI tool actually use it, and only 20 to 30 per cent of paid Copilot seats are touched in a given week.

In other words: the average business is paying for five seats to get one active user. If your finance team asked why you were paying for five phone lines and using one, there would be a meeting about it by Friday.

Why does a team ignore a tool it asked for?

You can have a Ferrari in the garage and still only ever drive it to the corner shop. The reasons are ordinary and human:

• The training was generic. A webinar about "AI in the workplace" tells nobody how to do Tuesday's actual job faster. People don't transfer abstract examples to their own work; someone has to do it with them once.
• One bad answer, written off forever. Someone asked a vague question, got a mediocre reply, and concluded the whole thing is hype. That first impression is now company folklore.
• Nobody knows what's allowed. Can I paste a client email in? A contract? Nobody has said, so the safe answer is to not touch it. Silence reads as no.
• There's no time to learn. The people who would benefit most are the busiest. Nobody saves time later by finding time they don't have now.
• Quiet fear. Some people worry that showing a machine can do part of their job is evidence against them. Unspoken, this is the strongest brake of all.

Notice what's not on the list: the tools. It's almost never the people either. It's that the tools were never fitted to the actual work.

What doesn't fix it

Mandates don't. "Everyone must use Copilot" produces logins, not usage, and quietly breeds resentment.

More tools don't. If the team isn't using the licence you have, a second subscription is just a second unused licence. Buying software is easy; changing Tuesday is hard.

And another all-hands demo doesn't. Watching someone else's example work is entertainment. Watching your own work happen faster is conversion.

What actually changes behaviour

Four things, in our experience, and none of them are complicated:

1. Train on their real work. Bring the team's actual documents, actual emails and actual reports into the room and run them through the tools live. When someone watches their own Friday-afternoon report happen in minutes, the argument is over.
2. One workflow per person. Don't ask people to "use AI more". Help each person nail one recurring task, the one they do every week, until it's habit. Habits spread sideways through a team faster than any policy.
3. Say what's allowed, in writing. A one-page policy: what can go into the tools, what can't, and who to ask. Permission unlocks more usage than any feature does.
4. Make the first win loud. When one person saves three hours a week, have them show the team how. Internal proof beats external experts every time.

Then measure the right thing. Logins tell you nothing; someone can open Copilot every morning and never let it touch real work. Ask each person one question a month: which tasks did AI take off your plate, and how long did they used to take? If the answer is "none" three months running, the rollout has stalled and needs one of the four fixes above, not another licence renewal. Most teams we work with find their first measurable win inside a fortnight once the training is built on their own tasks; if nothing has moved in a quarter, the problem is the approach, not the people.

When cancelling the licences is the right call

Sometimes the honest answer is fewer seats, not more training. If a role involves no writing, no email, no documents and no repetitive admin, an AI licence is a cost, not an investment; give those seats to the people drowning in text. And if you're paying for ChatGPT, Copilot and Gemini at the same time, pick one and consolidate. Duplicate tools split the learning and double the bill.

The point of the licences was never the licences. It was the hours. If a seat isn't returning hours, stop paying for it.

Where do you start?

Find out where the hours are actually going. Our two-minute quiz shows where AI is most likely to save time in your business, and emails you a personalised report. It takes less time than reading this article did.

Or book a call and talk it through. If the fix is a one-page policy and an afternoon of training rather than a consulting project, we'll say so. And if you're wondering whether outside help is worth it at all, here's what an AI consultant actually does, and when you don't need one.