Half of Gen Z workers in the UK feel guilty using AI to produce work, and 42% use it without telling their employer, according to research commissioned by employment software company Employment Hero. The same study finds that 64% of UK employers say AI has already changed what they look for when hiring, with AI capability now ranking among the top five qualities sought in candidates, ahead of prior experience. The gap between those two findings is not a curiosity. It is a live policy problem that the Chartered Institute of Personnel and Development (CIPD) has flagged repeatedly in its annual Reward Management Survey: hiring criteria are shifting faster than the workplace cultures that govern how employees are expected to work.

The numbers behind the shadow-skills problem

The Employment Hero study surveyed 1,025 UK business leaders and 1,500 UK workers as part of a wider piece of research covering more than 8,700 respondents across the UK, Australia, Canada and New Zealand. Within the UK Gen Z cohort, a coherent pattern emerges:

• 50% feel guilty using AI to produce work.
• 52% believe using AI to complete parts of their job amounts to cheating.
• 42% use AI at work without their employer's knowledge.
• 42% admit they often present AI-generated work as their own.
• 81% have taught themselves how to use AI through social media.

On the demand side of the same labour market, 37% of UK workers have already seen entry-level roles requiring AI knowledge, and 23% say they do not currently possess the AI skills needed to compete in an AI-driven job market. Employers continue to rank work ethic and attitude first, followed by communication and interpersonal skills, ability to learn quickly, and digital literacy. AI capability now sits inside that top five, ahead of prior experience.

Place those two panels side by side and the contradiction is hard to miss. Gen Z is self-teaching the skill employers say they need, then concealing the evidence of it. For HR practitioners who rely on the CIPD's Good Work Index as a benchmark for workforce engagement, this kind of hidden-skill dynamic is a signal that psychological safety in the workplace has not kept pace with technological change.

Why guilt is overtaking enthusiasm

The mood is not straightforwardly hostile to AI. 58% of Gen Z workers feel positive about AI becoming a bigger part of working life, against 25% who say they are worried. The friction arises when that optimism meets workplace culture and, in many cases, the absence of any formal policy guidance.

Kevin Fitzgerald, UK managing director at Employment Hero, described the core tension in the research write-up:

"There is a real contradiction emerging for young workers. They are being told that AI skills will be critical to their careers, and many are clearly enthusiastic about building those skills, but they still feel guilty when they actually use the tools."

Ria Kaur, a university student and jobseeker quoted in the same research, described how that tension plays out during internships and placements:

"As a Gen Z student experiencing the world of work through internships and placements, I see AI everywhere. But I also find that, in workplace situations, AI can feel like my dirty little secret."

Kaur attributed the secrecy to a specific generational stereotype:

"I think this comes from the stigma around younger generations using AI, which becomes stronger in the workplace because of the frustrating idea that Gen Z are lazy, or that we do not know what real work is. If a young person uses AI at work, it can feel like people assume they are offloading the task or taking the easy way out."

And she was clear that the actual usage rarely matches that caricature:

"In reality, a lot of us are using it responsibly to understand a task, prepare for a conversation or make sense of something new, but I still feel like it has to be kept hidden."

The ACAS (Advisory, Conciliation and Arbitration Service) framework for fair and transparent workplace communication is relevant here. Where employers have not published clear guidance on AI use, workers are left to infer norms from context, and that inference, in the absence of explicit signals, tends toward caution and concealment rather than openness. ACAS guidance on technology and employment stresses that ambiguity in workplace rules consistently produces inconsistent behaviour, not self-regulation.

For employees and candidates: using AI without hiding it

For anyone in the 42% using AI quietly at work, the risk is not only cultural. Presenting AI-generated output as your own work can breach an employer's acceptable-use policy where one exists. Beyond internal policy, there is also a UK GDPR dimension: under the Data Protection Act 2018, which gives domestic effect to the UK General Data Protection Regulation, inputting personal data, client information, or commercially sensitive material into a public AI tool is likely to constitute a data disclosure, potentially triggering obligations under Article 5(1)(f) (integrity and confidentiality) without any formal incident having occurred. The Information Commissioner's Office (ICO) has published guidance on generative AI and data protection, noting that organisations are responsible for what data flows through third-party systems on their behalf. If you are the person who pasted client data into a public model, you may be the source of that flow.

Practical steps for employees and candidates navigating this:

• Check the written rules before the unwritten ones. Ask your manager or HR whether there is an AI or generative-tools policy. If there is not one, that gap is informative. It usually means the employer has not yet decided, not that AI use is implicitly approved.
• Attribute your process, not just your output. Telling a colleague or manager that you used an AI tool to draft a first version, then rewrote the critical sections, is a professional disclosure, not a confession. It also protects you if the output is later questioned.
• Keep sensitive data out. Client details, personal data covered by the Data Protection Act 2018, unreleased financials, salary information: none of these belong in a public AI tool. Handling this boundary correctly is where responsible AI use becomes a demonstrable skill, not just a stated value.
• Prepare for AI-fluency questions in interviews. Employers are now asking candidates how they prompt, how they verify outputs, and what they would not delegate to a model. A real example from your own work experience carries far more weight than a rehearsed answer.

The reframing is straightforward, if uncomfortable: AI competence is now a hiring criterion. Concealing it does not protect your professional reputation. It strips your CV of the very evidence employers are screening for.

For HR leaders: the silence is a policy problem

If 42% of your Gen Z employees are using AI without your knowledge, and 42% are submitting AI-assisted work without disclosure, your organisation already has an AI programme, it just operates outside your governance framework. Three concrete implications follow.

1. A risk profile you cannot see. Shadow AI use creates shadow data exposure. If workers are passing internal information through public models to make sense of an unfamiliar task, there are no logs, no scope definition, and no defensible answer for a client or regulator asking whether their data has touched a third-party system. The ICO's guidance on generative AI is explicit that data-controller obligations apply regardless of whether the use was formally sanctioned. An HR or legal team that cannot account for AI data flows is operating with an unquantified liability.

2. A skills-development gap. 81% of Gen Z have taught themselves AI through social media. Their reference frame is influencer demonstrations and viral prompting tips, not your organisation's quality standards or data-handling obligations. Without structured guidance, the skill remains uneven and the productivity benefits employers anticipate stay theoretical. The CIPD's annual Reward Management Survey consistently shows that organisations linking learning and development investment to measurable capability outcomes outperform peers on retention at the 12-to-24-month mark, a horizon directly relevant for Gen Z joiners.

3. A hiring contradiction. 64% of employers say AI has changed what they hire for, and AI capability is now in the top five criteria. If new joiners rapidly learn that using AI openly is treated as an admission of corner-cutting, organisations have selected for a skill their culture then suppresses. That is a retention and engagement problem, and the ONS (Office for National Statistics) data on labour market flows suggests that workers aged 18-24 have among the shortest median job tenures in the UK workforce, making early culture signals disproportionately consequential.

Kevin Fitzgerald put the workplace-norms gap plainly:

"When half of Gen Z feel guilty using AI at work, and more than four in 10 are doing so without their employer's knowledge, it shows that workplace norms have not yet caught up with employee behaviour."

And on what a constructive response looks like:

"AI shouldn't feel like cheating. It should feel like using any other tool that helps people do their jobs better. But if workers don't have clear guidance, they'll continue to learn in the shadows, making it harder for businesses to understand AI's true impact, manage risk and support skills development."

The low-cost first moves for HR and people teams are those that convert shadow use into visible, governed use:

• Write the policy, even a short one. Which tools are approved for use, what categories of data cannot be entered, when to disclose AI involvement in a deliverable. A concise document, ideally co-developed with line managers and reviewed against ACAS guidelines on technology policies, removes the ambiguity that drives concealment.
• Make disclosure safe. The reaction a manager gives when a team member says they used an AI tool to draft something shapes the behaviour of the next hundred employees who witness it. If the response treats that disclosure as diligent and transparent, disclosure spreads. If it signals disapproval, secrecy spreads.
• Rework interview and onboarding scripts. If AI capability is in your top-five hiring criteria, ask about it directly at interview, and brief new joiners on day one about how the organisation expects AI to be used. DSIT's UK AI Opportunities Action Plan (published January 2026) explicitly calls on employers to embed AI literacy into onboarding as part of the broader national upskilling agenda.
• Fund the learning path. 23% of UK workers do not believe they currently possess the AI skills the job market now demands. If the training budget does not reflect that gap, self-taught social-media learning fills it by default, with no assurance of quality, accuracy, or alignment with data-handling obligations.

The wider signal for UK workplaces

The 50% guilt figure is striking, but it is a symptom rather than the condition itself. The underlying story is that hiring criteria have shifted faster than the workplace norms that govern how people are expected to work, and the cost of that gap is being borne disproportionately by younger workers. They are simultaneously being told that AI fluency is professionally essential and made to feel that demonstrating it is somehow dishonest.

For UK employers, the DSIT UK AI Opportunities Action Plan and the CIPD's ongoing work on responsible AI in people management both point in the same direction: organisations that build clear, fair frameworks for AI use, rather than leaving the question unanswered, are better placed to retain the workers they have invested in recruiting. The ICO's guidance on AI and data protection provides the compliance floor. The CIPD's Good Work Index and ACAS best-practice materials provide the cultural scaffolding above it.

Kevin Fitzgerald's closing observation captures what is at stake for employers willing to close the gap:

"The opportunity is to bring AI into the open, build trust, be transparent and help a generation use it confidently."

For employees, the practical implication is to stop treating a top-five hiring criterion like a secret. For employers, the implication is this: if you are selecting for AI capability, build the conditions in which people can actually demonstrate it.