Employees do not appear willing to wait for their employers to decide how generative AI fits into working life. They are already using it, often without formal approval, and some are paying for it themselves.
Deloitte GenAI Workforce Survey data puts scale behind that shift. The firm’s inaugural UK survey of 25,000 workers found that 63% knowingly use generative AI for work. It also found that 31% of GenAI users have used tools without their employer’s knowledge.
That figure will concern security and compliance leaders. It should also concern people leaders. Shadow AI is not only a question of which data may enter an unapproved tool. It can reveal a gap between what employees need to do their jobs and what their organizations are prepared to provide, explain, or discuss.
Deloitte’s research does not show that every unauthorized use is appropriate. It does show that many employees have moved ahead of their employers. Organizations now need to respond with more than a policy document. They need a clear employee experience strategy for AI.
TL;DR
Deloitte found that 63% of UK workers have used GenAI for work, but 31% of GenAI users have done so without their employer’s knowledge.
Workers reportedly spend an estimated £958 million a year of their own money on GenAI tools used for work.
Half of GenAI users have received no formal workplace training, while 65% do not see leaders discussing AI with good understanding.
For employers, shadow AI is a security issue, but it is also a trust, enablement, and employee engagement issue.
Why Is Shadow AI an Employee Engagement Issue, Not Just a Security Risk?
Shadow AI becomes an employee engagement issue when workers feel they need to solve everyday work problems outside the systems and support their employer provides. Deloitte found that nearly one in three GenAI users have used a tool without employer knowledge. That behavior can indicate risk, but it can also indicate unmet demand for faster, simpler ways to work.
A worker who uploads sensitive information to an unapproved chatbot creates a genuine governance concern. Yet an organization that treats every unauthorized use as misconduct may miss the underlying reason employees looked elsewhere in the first place.
Deloitte’s findings suggest this is not a marginal behavior. Among employees who use GenAI, 22% said they had used tools without their employer’s knowledge but believed their employer would approve. A further 9% said they believed their employer would not approve, while 4% reported using tools their employer had banned.
Those categories do not carry the same level of risk. They do, however, show that many employees make their own judgments about what work AI use is acceptable. That is a weak foundation for responsible adoption.
Workers also appear to use GenAI often enough for this issue to affect day-to-day work culture. Deloitte found that 24% of UK working adults use GenAI at least daily, while 12% use it multiple times a day. In HR and marketing roles, the daily-use figures reached 43% and 41%, respectively.
Organizations cannot manage a behavior that they only discover after it becomes routine. They need to understand which tasks drive employees toward external tools, what information workers are entering into them, and why approved alternatives have not met the same need.
What Does Deloitte’s £958 Million Personal AI Spending Estimate Reveal?
Deloitte’s estimate that UK workers spend £958 million each year on GenAI tools for work suggests employees see enough personal value in the technology to fund it themselves. The number does not prove that every subscription delivers a return. It does show that employer provision has not kept pace with at least some workers’ expectations.
The survey found that 17% of GenAI users pay for at least one external GenAI tool themselves. Deloitte calculated its annual £958 million estimate from self-reported monthly spending among those users.
Workers also use a mixed toolset. Forty-six percent of GenAI users said they use free tools at work. Thirty-four percent use external tools paid for by their company, while 17% use in-house tools.
What the Deloitte Survey Shows
Personal spending on work GenAI can signal employee initiative, but it can also expose uneven access to approved tools, training, and support. A workforce where some people self-fund premium capabilities may create a different experience for employees who cannot or will not do the same.
This creates an employee experience question that goes beyond expense reimbursement. Employees who pay for tools may feel empowered by their initiative. They may also feel that their employer has left them to handle a core work capability alone.
It can also create inconsistency. One employee may have access to a paid tool with stronger features, while another relies on a free version, an internal tool, or no tool at all. The result can be uneven productivity, uneven quality, and uneven exposure to risk.
Leaders should therefore avoid reading the £958 million estimate as a simple endorsement of bring-your-own-AI. It is a prompt to investigate which tools employees value, which workflows lack support, and what a fair approved access model should look like.
Why Are Employees Using AI Without Formal Training or Clear Permission?
Employees are using AI without formal training or explicit permission because access to public tools is easy while organizational guidance often remains incomplete. Deloitte found that 50% of GenAI users had received no formal training from their employer. That leaves many workers to decide for themselves how to use AI effectively and safely.
Deloitte’s findings echo recent UC Today reporting on Kyndryl’s People Readiness Report, which identified workforce preparation, governance, and work redesign as central factors in whether AI investments deliver business value.
The training gap sits alongside a leadership gap. Deloitte found that 65% of GenAI users did not perceive their leaders as talking about AI with a good understanding. This included workers who said leaders did not discuss AI at all and those who felt leaders discussed it without demonstrating sufficient understanding.
That combination invites informal behavior. Employees still have targets to meet, customers to serve, emails to write, information to find, and meetings to summarize. If they can access a tool that promises to speed up those tasks, many will try it.
Deloitte’s task data makes that pattern clear. GenAI users most commonly use tools for searching for information and writing emails, both at 43%. Summarizing followed at 31%, while 29% reported using AI for fact-checking.
These are not necessarily specialist or experimental activities. They are ordinary knowledge-work tasks. That fact makes the governance challenge more urgent, because employees may see GenAI as a practical utility rather than a technology that needs careful judgment.
Formal training should not only tell people what they cannot do. It should help them recognize safe use cases, understand when human review matters, protect sensitive data, challenge inaccurate outputs, and know which approved tools they can use for specific tasks.
Are Time Savings Improving Employee Experience or Simply Raising Expectations?
GenAI can improve employee experience when it removes repetitive work and gives people more capacity for valuable tasks. Deloitte’s data shows a more mixed picture. Half of GenAI users said the tools save them time, while the other half did not yet report a time benefit.
Across the full UK workforce, Deloitte found that 32% said GenAI saves them time. Among people who reported savings, the average figure was 70 minutes a week, according to Deloitte’s public release.
