UK Workers Just Put a Price on the AI Guidance Gap: £958 Million

Deloitte published the largest single-country study of workplace generative AI use so far this week, and the number that should stop most UK business owners mid-scroll isn't the adoption figure. It's the one about money.

The survey covered 25,000 UK workers. 63% have used generative AI at work, and 24% use it every day. Those numbers alone confirm something most leaders already suspect: the adoption question, whether people would actually use this stuff, has been settled for a while.

What's more revealing is what came next. 31% of the people using GenAI at work are doing it without their employer knowing. Almost 1 in 10 have used a tool their employer had explicitly banned. And 65% said their organisation hasn't given them anything close to convincing guidance on how GenAI should be used.

Put those three findings together, and a fourth number makes sense. Deloitte estimates British workers are now spending £958 million of their own money each year on GenAI tools for work. Not because the company refused to pay. Because nobody asked what they needed, so people went and solved it themselves, quietly, on a personal card.

It would be easy to read this as a discipline problem: staff going around policy, using banned tools, hiding what they're doing. That's the alarmist version, and it's the wrong one. A workforce willing to spend close to a billion pounds of its own money solving a problem isn't disengaged. It's telling you, as clearly as spending data ever tells you anything, exactly what it needs. The question is whether anyone in the business is positioned to listen.

This is precisely the gap the AI Accelerator Diagnostic is built to close, across two of its six dimensions in particular. People and Culture asks who in a business can already use AI well, and whether the organisation is actively supporting that or leaving it to chance. Governance asks what everyone needs to know before using AI in real work, so that support can exist without either a blanket ban or a free-for-all. Most businesses currently have answered neither question.

For a leader reading the Deloitte numbers and wondering what to do this week, the temptation is to reach for a policy document or a software restriction. Both are the wrong first move. A ban doesn't stop £958 million of demonstrated need, it just pushes it further underground, and a policy nobody helped write rarely survives contact with how people actually work.

The more useful first step is smaller and more honest: find out what's actually already happening. Which tools people are quietly relying on. What they're using them for. What would make that easier and safer to do openly, rather than on a personal subscription nobody in the business knows about. That conversation, done without judgement, tends to surface more useful information in an afternoon than any audit will.

It also tends to be the conversation people are relieved to have. Deloitte's 65% figure on unconvincing leadership isn't really a complaint about AI. It's a complaint about being left to work it out alone. Most employees using AI quietly aren't trying to get away with something. They're trying to do their job better, and they've concluded, reasonably, that asking permission first would take longer than just getting on with it.

None of this requires an enterprise-scale response. It requires someone in the business treating the £958 million figure as data rather than a scandal, and asking the two questions that follow from it: what has my team already worked out that I haven't caught up to, and what would it take to back that properly instead of pretending it isn't happening.

That's the conversation worth having before any policy gets written.

Go wisely.

Source: Deloitte GenAI Workforce Survey, reported by Bloomberg and TechRadar, 16 September 2026.