Nobody opens the tool.
The licences are paid for. Usage reports show a handful of logins in the first week and none after.
Home / AI training / When training didn't stick
Team training
Most AI training for teams is a half day of slides about the tools, a few generic prompts, and nothing anyone used the following week. If that's what you got, the problem wasn't your team. It was the session.
Run by the person who builds these systems. About Shanoor
Updated October 2026
In short
AI training fails to stick for three reasons: it teaches the tool instead of the job, it uses generic examples instead of the team's own work, and it skips the data and safety rules so nobody is sure what's allowed. Norloch's sessions are built on the team's real tasks, cover the tool the team actually has (ChatGPT, Claude or Microsoft 365 Copilot), end with a one-page usage policy, and are checked a month later. From £2,000 per team.
The difference
| What you probably got | Why it didn't stick | What we do |
|---|---|---|
| Slides about how AI works | Nobody needs the theory to write a better email | Ten minutes on how the tools work, then hands on |
| Generic prompt examples | Your team doesn't write poems or marketing copy | Exercises built from your own inbox, documents and processes |
| One tool, usually ChatGPT | Your team has Copilot, or Claude, or all three | Whichever your team uses, and when to reach for which |
| Nothing about data | So people either paste client data or stop using the tool | Plan settings checked together; a one-page usage policy to keep |
| A certificate | Proves attendance, not change | A priced list of what's worth automating, and a check a month later |
Signs it didn't stick
The licences are paid for. Usage reports show a handful of logins in the first week and none after.
"Is it safe to put this in?" and "what would I even use it for?" are still being asked a month later.
Usually the person who already used it before the training.
FAQ
Because the first one almost certainly taught the tool and not the job. A session built on your own emails, documents and processes gives people something to do on Monday morning. The test is simple: at the end, can each person name one task they'll do differently tomorrow?
Yes. Copilot, ChatGPT and Claude, and the differences between them, including the data settings that matter on each.
From £2,000 per team, scoped on a free call. If we don't think a second session will stick either, we'll say so.
We ask for three tasks the team does every week before the session, and check a month later whether they're doing them differently. If not, the session hasn't worked and we want to know.
Free 30-minute call. Tell us what you got last time. We'll tell you honestly whether a second session will land.
Related