You draft a client email with an AI assistant, polish the wording yourself, press send and mention it to nobody. If that sounds familiar, you are far from alone, particularly if you were born after 1997. Research from Imperial Lifelong Learning, a new division of Imperial College London, puts a precise figure on how common this habit is, and the number carries a practical warning for anyone managing their own career.

The study found that 29% of Gen Z workers use AI without their employer's knowledge, compared with 18% of millennials. The report labels this pattern shadow AI, and the reasons behind it reflect something important: the gap between how capable younger workers actually are with these tools and how visible that capability is to the people making decisions about their pay, projects and progression.

What the Imperial College research actually found

The survey covered 2,000 UK workers across five industries. Gen Z here means people born between 1997 and 2012, so the group spans those in their first roles to professionals approaching thirty. The generational breakdown on regular AI use is not quite what you might expect.

Nearly half of Gen Z respondents, 48%, use AI regularly at work. That places them behind millennials at 55%, but ahead of Gen X at 45% and baby boomers at 31%. In other words, younger workers are not the heaviest users by frequency. They are, however, the generation most likely to keep it to themselves, and that gap between usage and disclosure is where the career risk sits.

Heather Haseley, chief lifelong learning officer at Imperial Lifelong Learning, framed the broader context clearly: AI is transforming high-growth industries and promises to create new opportunities to accelerate productivity and economic growth. The concern, she added, is that few technologies have been adopted by businesses this rapidly, creating a real risk that workforce skills fall behind what AI can help people achieve.

Why so many younger workers keep quiet

The research points to stigma rather than deliberate rule-breaking. Among Gen Z respondents, 35% said using AI feels dishonest, and 26% believe colleagues would view them negatively for it. Those are substantial numbers, and they suggest that secrecy is often a response to perceived social risk rather than clear policy violations.

Part of that anxiety has roots in education. Industry leaders interviewed for the Imperial research described a disconnect between how AI is treated in the workplace and how it is handled in formal learning environments, where young people may be penalised for using it. That uncertainty, as Haseley noted, follows them directly into work, and higher education needs to bridge that gap.

Unclear workplace policies compound the problem. The Advisory, Conciliation and Arbitration Service, known as ACAS, has published guidance encouraging employers to involve staff in developing AI policies and to communicate expectations clearly. When that guidance goes unheeded and nobody has told you whether a tool is permitted, saying nothing can feel safer than raising your hand. Silence, though, carries costs of its own.

The CIPD, the Chartered Institute of Personnel and Development, whose annual Reward Management Survey tracks workforce development trends across the UK, has consistently noted that skills visibility is as important as skills acquisition in performance and pay conversations. If your manager cannot see how you work, they cannot reward what you produce.

What hiding your AI use can cost you

Shadow AI can feel like a private productivity shortcut. In practice, it tends to work against you in three distinct ways.

• Your skills stay invisible. If your output is attributed to speed or natural ability alone, nobody thinks to offer you AI-related projects, training or additional responsibility. The skill exists; the opportunity to build on it does not follow.
• You carry the compliance risk alone. Pasting client details, internal figures or personal data into a tool your employer has not approved can breach your organisation's data policies and, more seriously, engage obligations under the UK GDPR as administered by the Information Commissioner's Office (ICO). If something goes wrong, you have no cover and no paper trail showing you acted within sanctioned boundaries.
• You miss structured learning. Trial and error in private rarely comes with the feedback that turns a habit into a provable, transferable skill. Formal programmes tied to approved tools do.

The wider picture adds weight to these individual risks. Across all generations surveyed, 59% of workers believe there is a skills gap in their industry, 66% think the UK's international competitiveness will suffer if it is not addressed, and 70% expect that failure to affect the jobs available to the next generation within five years. Haseley put the pace of change bluntly:

Research shows professional skills now have a shelf life of around five years, less for technical skills.

The Office for National Statistics (ONS) has found that the share of businesses using at least one AI technology almost tripled between 2023 and 2026. The UK government has predicted AI adoption could boost the economy by up to £400bn by 2030. A skill that organisations are actively investing in is a skill worth being seen to have. Staying silent about it means your employer is making those investment decisions without knowing what their workforce can already do.

How to bring your AI use into the open

This does not require a confession or a lengthy conversation. A few measured steps can shift a quiet habit into something that actively supports your career.

• Find the policy first. Check your staff handbook, intranet or IT guidance for anything covering AI tools. Many larger UK organisations have now published lists of approved tools following guidance from ACAS and sector-specific regulators. If a list exists, sticking to it removes the ambiguity entirely.
• Ask with a concrete example. Framing a question around something specific, such as explaining that you have been using an AI tool to draft first versions of meeting notes before reviewing and correcting them yourself, and asking whether that is acceptable and whether a preferred tool exists, lands far better than a vague enquiry about whether AI is permitted.
• Keep sensitive data out. Until you have an explicit green light, do not put client names, colleague details, financial figures or any personal data into an external AI tool. The ICO's guidance on AI and data protection is clear that individuals can carry personal liability when data is processed outside agreed channels.
• Show your checking. When you share work, explain what you verified, corrected or rewrote. The judgement you applied is what your line manager values, and demonstrating it directly addresses the concern, common in the Imperial findings, that using AI is somehow dishonest.
• Ask for formal training. If your organisation is rolling out AI tools, ask whether there is a programme you can join. For roles in sectors covered by the DSIT UK Semiconductor and AI strategy, employer-funded upskilling is increasingly a live option. Asking signals initiative rather than shortcut-seeking.

If the answer is a firm prohibition, that is useful information in its own right. It tells you something concrete about how the organisation treats emerging skills, and it is worth factoring in the next time you read employer reviews or weigh up a job offer.

Talking about AI skills when you apply for roles

Hiring expectations are shifting alongside workplace practice. In a separate global study published in July 2026, Employment Hero found that 37% of UK workers say entry-level roles now specify AI knowledge as a requirement. Being able to describe how you use AI responsibly is becoming a standard part of the job hunt rather than a differentiator.

Prepare one specific example for interview: the task, the tool, what you checked and what you changed before submitting. On your CV, pair each AI tool with the outcome it helped you deliver rather than listing it as a bare keyword. A line noting that you used an AI drafting tool to reduce first-draft time by half, then edited for accuracy and tone, is more credible and more useful to a recruiter than the word AI sitting alone under a skills header.

It is also reasonable to ask the interviewer how the team uses AI and whether the organisation has a written policy covering it. The answer tells you something important: whether you would actually be able to use your skills openly once you join. Organisations that have thought this through tend to have better-defined answers and, according to CIPD commentary on workforce capability surveys, stronger retention among technically capable younger staff.

The wider signal for your career

The headline figure from the Imperial Lifelong Learning research is about secrecy, but the practical lesson is about visibility. Haseley's finding that trepidation around AI use is widespread among UK workers reflects a moment of genuine transition: the tools are ahead of the norms, and workers who have already developed real AI capability are carrying an asset their employers cannot currently see or measure.

For you, the opportunity is straightforward. If you already use AI well, your next step is not to use it differently, but to make sure the people who decide your projects, training and pay can actually see it. The ONS data on business AI adoption and the government's £400bn economic forecast both point in the same direction: this skill is becoming foundational, not optional. Keeping it hidden means you are doing the work without getting the credit, and that is a cost worth eliminating.