What UK HR Leaders Asked About AI (And the Answers That Actually Matter)

It's hard to escape the conversations around AI at work. From productivity to new features within existing platforms, it has already become part of how businesses run day-to-day. For HR leaders, that raises some questions.

Early on, you might have been asking: What can AI do? Now, the more pressing questions are about what kind of adoption HR should spearhead, concerns about employee data, bias, and trust, and whether AI is actually improving the way people work. All of which are valid! 

A CIPD poll found that 63% of people would trust AI to inform an important workplace decision, but only 1% would trust it to make that decision. 

That is a useful starting point for HR. As it stands, AI can support people and processes, but responsibility should still always sit with human insight.

So, what are UK HR leaders actually asking about AI right now? And what answers can help them move forward with more confidence? We explore some common questions below!

“Should we be using AI, and where should HR start?”

The best place to start is not with the tool, but with tackling a genuine problem that needs your attention.

AI adoption can quickly become overwhelming when every platform claims to have a new feature, every vendor promises efficiency, and every department wants to experiment. But HR does not need to roll AI out across every process immediately.

A better first question is: where are employees, managers or HR teams experiencing unnecessary friction? It is most likely a repetitive process, like a delay in finding information, a recurring question that takes up HR time or a workflow where people are doing the same manual task again and again. 

The first AI project should be:

• Narrow enough to manage
• Easy to compare against the existing process
• Low risk if the output is wrong
• Reversible if it does not work
• Have a clear owner

Among businesses already using AI, UK Government research found that 30% of staff currently use it on average. Just over half of businesses using AI said they use it constantly. That shows AI is already becoming part of everyday work for many organisations, but it does not mean every HR team needs to move at the same pace.

Instead of creating a long list of possible use cases, define:

• One problem
• One workflow
• One employee group
• One accountable owner
• One measurable outcome

Avoid starting with decisions that directly affect someone’s employment, pay, performance or access to opportunities. Those areas carry higher risks and need more careful governance.

For HR teams still exploring practical applications, low-risk use cases such as policy Q&A, onboarding support, benefits guidance or HR communications can be a sensible place to begin. The goal is not to launch AI everywhere, but to start learning where it genuinely helps before scaling.

“Can we safely put employee data into an AI tool?”

This is one of the biggest questions for HR, and rightly so! HR is privy to some of the most sensitive information in an organisation, including health, absence, performance, pay and demographic data. 

That does not mean employee data can never be used with AI. It does mean HR needs to understand exactly what is happening to that data.

There is an important distinction between a public generative AI tool, an enterprise-approved AI platform, and AI already embedded within an HR system. They may look similar to the user, but they can have very different rules around storage, access, retention and model training.

Before putting employee data into any AI tool, HR should be able to answer:

• What data is being entered?
• Why is it needed?
• Where is it stored?
• How long is it retained?
• Who can access it?
• Can it be deleted?
• Is it used to train the provider’s models?

One audit found that some tools collected more personal information than necessary and retained it indefinitely. Which is why Acas advises against entering personal or commercially sensitive information into public AI tools without checking company policies and approved platforms.

For HR, the practical answer is not “never use employee data”. It is to use the minimum information needed, ensure the organisation understands the data flow, and involve legal, data protection, and IT security teams before introducing tools that process sensitive employee or candidate information.

In simple terms, if the organisation would not be comfortable explaining its use of the data to an employee, it is probably not fit for use.

“How do we know the AI is not biased?”

The simple answer is: you don't! It requires rigorous tests, questions and monitoring, and still it could generate incorrect information. 

Not only that, but AI also confidently get things wrong. The language it uses gives the impression that it is objective, but that output may still reflect patterns in historical data, assumptions built into the system, or proxy variables that indirectly disadvantage certain groups.

Bias can appear through all sorts of channels, including:

• The information used to train the system
• The criteria chosen by the organisation
• The way an algorithm weighs information
• The way managers interpret and act on an output

As a tool programmed and managed by people, it will unintentionally have bias. This becomes especially important in recruitment, promotion, performance, pay, absence or redundancy-related processes. These are areas where an AI-generated output could influence a decision with real consequences for someone’s working life.

The ICO found examples of AI recruitment tools that allowed recruiters to filter out candidates with certain protected characteristics. It also found tools inferring characteristics such as gender and ethnicity from a candidate’s name, rather than collecting accurate information directly.

That is why HR should not rely on a vendor’s claim that a system is fair or objective and should keep human decision-making as the last line of defence.

When trying to understand the reliability and bias of an AI tool, you can ask:

• What data was the system trained and tested on?
• Has it been tested across different demographic groups?
• What differences in outcomes have been identified?
• Can the provider explain what influenced the output?
• What happens when bias is found?

Human oversight matters, but only if it is meaningful. If a manager simply accepts an AI recommendation without understanding how it was reached, that is not really oversight. The reviewer needs enough information, authority and time to challenge the output.

Candidates and employees should also know when AI is being used and how they can question an outcome. AI can support people's decisions, but it should not quietly become the decision-maker.

“Who needs to be involved before we introduce it?”

AI adoption should not be treated as an HR-only or technology-only project.

For many UK employers, the right conversation will involve recognised trade unions, employee forums or other worker representatives. Works councils may not be the default structure for every UK-only employer, but multinational organisations with existing European consultation arrangements should still consider them.

The broader question is: 

• Who will use the tool?
• Who will be affected by it ?
• Who needs to be involved before decisions are made?

That might include HR, legal, data protection, IT and information security, procurement, and relevant managers and employees who will experience the change day-to-day.

Consultation becomes especially important where AI affects how employees are monitored, how work is scheduled or allocated, how recruitment or performance is assessed, or how job responsibilities may change.

The timing matters. Employees should be involved while the organisation is still defining the problem and selecting the system, not only after the tool has already been purchased.

That early input can reveal risks leaders may not see from the top. It can also make adoption more realistic because employees are more likely to trust a tool when they understand why it is being introduced and how it will affect their work.

“Why are employees not adopting the tools we have introduced?”

Access does not equal adoption; we know the same holds true for wellbeing initiatives. Both require a transitional period and ongoing upkeep. 

Employees may have an AI tool available, but that does not mean they know when to use it, how to use it well, or whether using it will be seen positively. 

Employees may also struggle with the idea that:

• AI could replace parts of their role or their entire position
• Using AI could make their own contribution appear less valuable
• Their prompts or activities are being monitored
• An incorrect output could be held against them
• They will be expected to complete more work in the same amount of time

Those concerns are not hypothetical. An Acas survey found that job losses were workers’ most common AI concern, cited by 26%. A further 17% were most concerned about errors, and 15% about insufficient regulation.

That means adoption is not just a training issue. It is a trust issue, a communication issue and a work-design issue.

Employees need clear guidance on approved tools, examples that relate to their role and time to learn without feeling like they are being tested. They also need clarity on what remains a human responsibility and how to report something that looks wrong.

For employees to feel safe using AI, they need:

• Clear guidance about approved tools
• Examples relevant to their role
• Time to learn and experiment
• Clarity on what remains a human responsibility
• A safe route for reporting problems

The good news is that employees can respond positively to AI when it removes frustrating work. The CIPD has reported that, among employees who experienced tasks being automated by AI, 85% said it improved their performance, with positive associations also reported for job satisfaction and mental health.

That is the key distinction, AI adoption is easier when employees can see how it removes friction, rather than simply increasing expectations.

Managers have a major role here. They need to demonstrate responsible use, explain how AI should and should not be used, and be clear about what any time saved is meant to achieve.

“What should we measure?”

Theinstinct is to measure logins, licences, tokens spent and prompts entered to show activity, but they do not, by any measure, prove AI is improving work.

Before launching an AI tool, HR should establish a baseline. Only then can HR tell whether AI is making a meaningful difference. Useful efficiency measures might include:

• Time needed to complete a process
• Employee or manager waiting time
• Number of cases handled
• Repetitive administration reduced.

But efficiency is only one part of the picture. HR should also measure quality, like:

• Are outputs accurate, useful and consistent?
• How often do they need to be rewritten or corrected?
• Are employees and managers satisfied with the result?

Then, risks need to be measured too, to weigh against the gains:

• Incorrect or fabricated outputs
• Human overrides
• Complaints or challenges
• Differences between demographic groups
• Data protection or security incidents

Adoption should focus on repeat use within the intended workflow, not on whether someone opened the tool once.

Government research found that 56% of businesses using AI reported an increase in employees’ overall productivity since adopting it, while 35% reported no change and 1% reported a decrease. The same report notes that these are self-reported figures, so they should be treated as estimates rather than definitive proof of impact.

That is why HR should be cautious about broad claims based only on estimated hours saved. A better question is: what happened to the time?

• Was it used for better employee support?
• Did it create more capacity for strategic work?
• Did it reduce overtime or pressure?
• Or were employees simply given more tasks?

If AI makes a process faster but leaves people feeling more stretched, it has not solved the right problem.

“Is AI making work better, or only faster?”

This may be the most important question for HR, because AI can remove repetitive work, reduce frustration and give people more time for judgement, relationships and creative thinking.

Used well, it can help HR teams spend less time searching, sorting and drafting, and more time supporting people. But AI can also increase workload, monitoring and the expected pace of work if every efficiency gain is treated as extra capacity.

That is where the conversation needs to connect with employee wellbeing. The Work-Life Wellness Report 2026 found that 90% of employees experienced burnout symptoms in the previous year. Wellhub’s research also found that 89% of employees say they perform better when they prioritise their wellbeing, while 86% consider wellbeing at work as important as salary. 

This gives HR another way to evaluate AI success. Alongside speed, cost and productivity, organisations should look at whether AI affects:

• Perceived workload
• Stress
• Autonomy
• After-hours work
• Ability to take breaks
• Time available for meaningful work and wellbeing

Because saving time only creates value when organisations make deliberate choices about how that time is used.

If AI helps an HR team answer policy questions faster, that adds values. If it helps managers prepare for better conversations about workload, that moves the needle. If it helps employees find the right support before stress becomes burnout, that can go along way for a healthier employee experience.

But if AI simply makes it easier to ask more of already stretched teams, it risks becoming a faster route to burnout.

The answer that matters most

AI has real potential for HR, but introducing it well is not just about choosing the right technology, instead, leaders must approach it with care and first grapple with:

• Understanding how employee data is processed
• Demanding evidence about bias
• Involving employees and their representatives
• Addressing the real barriers to adoption
• Starting with a defined problem
• Measuring quality, trust and wellbeing alongside productivity

This way, AI can create more capacity for the human side of HR, but that benefit should not be assumed. It has to be designed, communicated and measured.

The organisations that get the most from AI will not necessarily be the ones that automate the most. They will be the ones who use AI to remove friction while protecting what employees still need most: trust, support, clarity and time to care for their wellbeing.

Technology can help make work easier. But employees still need a culture that elevates them, which can include an accessible, flexible ways to support their physical, mental and everyday wellbeing.

That is where Wellhub can help. By bringing fitness, mental health, nutrition, sleep and holistic wellbeing support into one flexible platform, Wellhub helps organisations make wellbeing easier to access and easier to sustain.

Build a healthier, more wholesome employee experience with Wellhub.