Personio reports that 42% of UK employees use AI in their role, with some use happening without employers’ knowledge. Its Workforce Pulse release was published on 15 October.

Censuswide surveyed 6,000 employees and 3,001 HR decision-makers across the UK, Germany, Spain and the Netherlands. The fieldwork took place between 29 May and 10 June 2026.

Personio sells HR software and commissioned the research. The sample covers workplaces across those countries, so you’ll need your own evidence to understand what’s happening in your organisation.

I’d start with the work people are doing. Asking whether they use AI responsibly invites an answer that tells you very little.

Ask about an actual task

Ask a colleague to describe the last task they used AI for. Which tool did they use, and was it through a work account or a personal one?

Find out what information went in. Then ask how they checked the result before using it.

You might hear about a generic parent-evening reminder drafted from public information. You might also hear about a file containing individual family circumstances being uploaded to the same tool.

Those activities need different responses. The name of the product alone won’t tell you whether the work fits your organisation’s rules.

A manager can be unaware of an appropriate use of an approved tool. Equally, a familiar tool can be used with information it shouldn’t receive.

Explain what you’ll do with the answers

Before asking staff about their use, explain who will see their responses and what will happen if they raise a concern. Only promise anonymity if you can provide it.

Be clear about the route for reporting a possible data incident. Someone who thinks they’ve made a mistake needs to know who to contact promptly.

Alongside that, make room for ordinary questions. A colleague should be able to ask whether a task is suitable without having to interpret a supplier contract first.

Put the guidance beside the work

Once you know which tasks keep coming up, write a few examples staff can use. For each example, explain which information is permitted and what needs checking in the output.

Put the guidance where colleagues will find it when they’re working. Test that with someone who didn’t help write it.

Give the guidance a named owner. Ask them to review the questions staff raise over the next month and fix any instructions that keep causing confusion.

Check the practical results too. Can colleagues complete the task using the approved route, or are account problems and missing permissions sending them elsewhere?

The survey can’t tell you which response will work in your setting. You can find out by following up with the same people and checking whether their difficulties have been resolved.

If a colleague found a useful AI tool this afternoon, would they know how to ask whether they could use it?

Source