British workers have embraced generative AI faster than almost anyone expected — and many of them are paying for it out of their own pockets. But according to the largest survey of its kind, the productivity payoff remains stubbornly small. Deloitte's inaugural GenAI Workforce Survey, published in mid-September and based on responses from 25,000 UK workers polled by Ipsos, paints a picture of enthusiasm outrunning impact: two-thirds of workers have tried AI tools, yet around half of those users say the technology saves them no time at all.
The findings, which have been dissected across the business press through late September, amount to a reality check for the AI productivity story. Adoption is no longer the question. What workers actually do with the tools — and what their employers do to help — appears to matter far more.
The headline numbers
The survey found that roughly two-thirds of UK workers have tried AI tools such as ChatGPT, Claude, Google Gemini and Microsoft Copilot, and nearly a quarter are using them daily in their jobs. Workers are adopting AI faster than their employers are making tools available, the responses suggested.
But the time savings are modest. On average, workers say AI saves them about 70 minutes a week. Only around a quarter of staff report time savings of more than two hours per week, and just 10% say they save more than five hours a week, according to analysis of the survey data. More than a third of AI users say they get no time back from using it at all.
The gap between usage and benefit is the survey's central finding. As one economist noted in response to the results, adoption does not automatically equal productivity — the impact is more likely to depend on how organizations invest in skills, training and new ways of working.
Shadow AI: a third use it in secret, a sixth pay for it themselves
Perhaps the most striking finding is how much of this adoption is happening outside official channels. Some 31% of people surveyed are using generative AI without their employer's knowledge — so-called shadow AI — and 17% are paying to use at least one AI tool themselves for work. One in six workers in Britain is paying for an AI subscription to help do their job, at a total estimated cost of nearly £1 billion ($1.4 billion) a year of their own money.
"UK workers are showing they don't want to wait for permission to use GenAI," said Deloitte UK's chief AI officer Hayley McKelvey. "Many are already using free tools or pay for premium versions themselves to help them get work done."
The security implications are significant. Company documents pasted into a personal AI account sit outside the firm's security controls, and personal logins and API keys are targets in their own right. Yet only 28% of organizations have clear AI policies, according to the research — meaning most employers have neither approved the tools their staff are using nor set rules for using them safely.
What workers actually use AI for
The survey suggests the modest time savings have a straightforward explanation: most people are using AI for relatively basic tasks. The three most popular uses are searching for information (43%), drafting emails (43%) and writing summaries (31%).
As analysts have pointed out, none of those tasks dominates most people's working lives — unless your entire job is finessing emails. The deployment of AI, at least in the UK workforce surveyed, remains surface-level: people are using it as a faster search engine with some drafting capabilities attached, rather than as a tool woven into complex work.
"Yet for many workers its role remains relatively basic, focused on searching for information and drafting content rather than supporting more complex work," said Deloitte partner Paul Lee. The deeper productivity gains — automating multi-step workflows, analyzing data, restructuring how work gets done — appear to be reaching only a minority so far.
The stigma problem
The survey uncovered a cultural obstacle that may be suppressing both honest use and real gains. Nearly a quarter of workers (23%) think there is a stigma attached to using AI at work, and 64% of weekly users worry their managers will conclude that AI can do their jobs. Deloitte found that people who feel that stigma are more likely to hide their use — feeding the shadow-AI problem from the other direction.
The dynamic is self-reinforcing: workers hide their AI use because they fear being judged or replaced, employers cannot see how AI is actually being used, and so no training, guardrails or workflow redesign follows. The productivity that might come from open, skilled use never materializes.
Why the gains are small — and what could change them
Three structural gaps run through the findings. First, training: about half of AI users say they have had no formal training on using the tools safely and well. People are improvising with powerful technology and mostly landing on the simplest use cases. Second, guidance: with only 28% of organizations having clear AI policies, most workers are guessing at what is allowed. Third, purpose: few employers appear to have redesigned workflows around AI; the tools are layered on top of unchanged jobs.
"The story here isn't that workers are using GenAI," said Paul Lee, Deloitte UK's partner and head of industry insight. "It's that they are using it despite limited training, patchy guidance and, in some cases, without their employer's knowledge."
Deloitte's prescription follows directly: "The organisations that gain the greatest advantage from AI will be those that move beyond simply providing access to tools and focus on helping people use them with the appropriate guardrails and purpose," Lee said. In other words, the next phase of AI productivity is not about licenses — it is about literacy.
Deloitte's chief AI officer made the leadership challenge blunt in a video discussing the findings: "The question for leaders isn't whether AI is being used. It's whether their organisations are keeping up."
What this means for workers and employers
For individual workers, the survey offers a useful calibration. If AI is currently saving you an hour a week on emails and summaries, you are squarely in the mainstream — and the bigger gains likely lie in learning to apply the tools to harder, more specific parts of your job, where training and experimentation pay off. The stigma findings also suggest a practical step: having an explicit conversation with your manager about approved AI use can be more productive than quiet improvisation.
For employers, the message is harder to ignore. A third of staff using AI in secret, a sixth paying for it themselves, half the user base seeing no time saved — these are symptoms of a strategy vacuum, not a technology failure. The firms that provide approved tools, real training and clear rules stand to capture both the productivity and the security upside. The rest, as one analysis put it, are relying on personal subscriptions they cannot see.
Deloitte says it will repeat the survey about every six months, which will make it one of the better longitudinal measures of whether the AI productivity story is maturing — or stalling. For now, the lesson of late September 2026 is clear: giving everyone a chatbot was the easy part. Teaching organizations to work with it is the real productivity project.
Sources
