How to Use AI in Employee Wellbeing Without Crossing the Line

HR teams like yours may be exploring how AI could make wellbeing support more relevant, accessible, and easier to navigate.

Used thoughtfully, it could help employees find options that better suit their goals, routines and preferences! But because wellbeing data is sensitive, personalisation can quickly feel intrusive if people do not understand how it works, what information is being used or what their employer can see.

The challenge is to improve support without crossing into monitoring. That means getting transparency, consent, governance and the “personalisation vs surveillance” line right from the start.

How can HR use AI to support employee wellbeing?

AI can make wellbeing programmes easier to access, more relevant and more inclusive, without requiring employers to track or assess individual employees. In practice, this might include:

• Helping employees discover wellbeing options that match their interests, such as fitness classes, mindfulness sessions or nutrition support
• Personalising how wellbeing content is presented based on preferences employees choose to share
• Sending optional, personalised reminders or nudges
• Improving search and navigation so employees canfind support that fits their schedule, location or goals
• Giving HR aggregated insights into which types of support are most popular, helping teams improve the overall programme

Importantly, these uses should focus on supporting employee choice rather than analysing behaviour or predicting outcomes.

For example, AI could help an employee find a nearby fitness class at a convenient time. It should not use that employee’s activity to assess their health, motivation or likely performance at work.

HR should not use AI to diagnose health conditions, assess wellbeing risk at an individual level or make decisions about performance or employment.

The goal is to remove friction and make support easier to access, not to monitor how employees use it.

The line between personalisation and surveillance

The difference between personalisation and surveillance comes down to who the technology is designed to serve.

Personalisation helps the employee make a better-informed choice. Surveillance helps the employer form a judgement about the employee.

Trust in the technology will also depend on the wider workplace culture. According to the Wellhub Work-Life Wellness Report 2026, only 17% of UK employees strongly agree that wellness is ingrained in their company’s culture.

Where wellbeing can sometimes feel bolted-on or performative, employees may be more likely to interpret AI-generated recommendations as another form of monitoring rather than genuine support. Personalisation should therefore be:

• Visible
• Voluntary
• Employee-led
• Easy to change or switch off

It starts to resemble surveillance when data is collected passively or unexpectedly. It also becomes more concerning when AI infers stress, burnout or mental health conditions from an employee’s activity.

Other warning signs include individual wellbeing activity being visible to managers or being used to influence decisions about performance, promotion, absence or disciplinary action.

A practical test can help HR stay on the right side of the boundary: Is the system supporting the employee, or assessing them?

Transparency: employees should understand how personalisation works

Transparency is essential when AI is being used with personal wellbeing information. Employees should be privy to:

• What information is being collected
• Why it is needed
• How AI is using it
• Who can access the results
• Whether individual activity is visible to the employer
• How long the information is retained

These explanations should be clear and easy to understand. Broad statements about using data to “improve the employee experience” are unlikely to give people enough information to make an informed decision.

HR should explain the specific use case. For example, an employee might be told that their selected interests will be used to recommend relevant wellbeing activities.

It is equally important to explain what the system does not do. Employees may need reassurance that individual searches will not be shared with HR, that managers cannot see which resources they access, and that the system does not create health-risk scores for the employer to review.

Research from the Institute for the Future of Workfound an association between greater exposure to AI, machine-learning tools and wearables and lower quality of life among UK workers.

The same research found positive associations between quality of life, employees’ perceived rights at work and HR approaches that emphasise wellbeing.

This suggests that the impact of AI depends on more than the technology itself. Agency, protection and employee involvement all influence how workplace technology is experienced.

Consent: personalisation should be a genuine choice

Employees should be able to access core wellbeing support without having to agree to extensive personalisation. Consent should be:

• Specific to the information being used
• Clearly separate from general platform access
• Easy to withdraw
• Free from negative consequences

Employees should also have control over how their information is used. For example, they should be able to choose:

• Which preferences inform their recommendations
• Whether previous activity is used
• Whether they receive personalised nudges
• Whether their activity contributes to aggregated reporting

These choices should not be bundled together. An employee may be comfortable receiving recommendations based on interests they have selected, but not want previous activity to influence future suggestions.

They should also be able to change their mind. Turning off personalisation or withdrawing consent should be simple and should not mean losing access to meaningful wellbeing support.

Personalisation should be an optional way to improve the experience, not a condition of accessing it.

Governance: agree the boundaries before launching the technology

Strong governance helps HR define where AI can be useful, where additional safeguards are needed and where the technology should not be used at all.

These boundaries should be agreed before a tool is introduced. Governance should involve HR, legal, data protection, IT and security teams, alongside employee representatives where relevant. Together, they should consider:

• Which data the tool can use
• Whether it can make sensitive health or behavioural inferences
• What the employer can see at an individual level
• Whether data could be used for any employment-related decision
• How employees can correct or delete their information
• How recommendations are tested for bias, accuracy and harm
• When human review is required

Employer reporting should default to aggregated or anonymised insight wherever possible.

This can help HR understand overall engagement, identify popular types of support and improve the programme without revealing what individual employees are doing.

Wellbeing data should also remain separate from performance management and workforce risk scoring. An employee’s use of sleep, stress, or mindfulness support should remain private.

These questions sit within a much wider conversation about responsible AI at work. For more on data protection, bias, employee trust and where HR should begin, read our guide to the AI questions UK HR leaders are asking – and the answers that actually matter

A practical framework for assessing AI wellbeing features

Not every use of AI carries the same level of risk. A simple framework can help HR decide which features may be appropriate and which require greater scrutiny.

Lower risk

Lower-risk uses are generally based on information employees actively choose to provide. Including:

• Recommendations based on employee-selected interests
• Personalised search and navigation
• Adjustable communication preferences
• Aggregated programme insight

The purpose of these features is to make support easier to find, rather than to draw conclusions about an individual employee.

Requires closer scrutiny

Some features may still support employees, but involve more sensitive information or more active interpretation. These could include:

• Recommendations based on past usage
• Automated wellbeing nudges
• AI chatbots handling sensitive conversations
• Tools that prioritise or triage support

HR should understand how these features work, what information they use and what happens when the technology makes an inaccurate or inappropriate recommendation.

Employees should also be able to understand, change or challenge how these features affect their experience.

High risk

High-risk uses move beyond helping employees access support and towards assessing their health, behaviour or employment risk. These include:

• Individual burnout or mental health scores
• Emotion or mood detection
• Manager alerts based on wellbeing activity
• Predictions about absence, performance or retention
• Personalisation that employees cannot understand or disable

These features are much more likely to cross the personalisation vs surveillance line and undermine employee trust.

Where Wellhub fits

AI can make wellbeing support more relevant and easier to use. But relevance cannot come at the expense of employee trust.

Transparency helps employees understand how the experience works. Consent gives them meaningful control. Governance prevents personalisation from expanding into inappropriate monitoring.

One of the most effective forms of personalisation is also one of the simplest: giving employees meaningful choice.

Different employees need different types of support, and those needs may change over time. Someone may be focused on fitness today, sleep next month and mindfulness during a particularly demanding period.

Wellhub gives employees access to all-encompassing wellbeing options across fitness, mindfulness, nutrition, sleep and everyday wellbeing.

AI can help employees navigate those options more easily without requiring employers to build detailed individual health profiles. The focus remains on helping people find what works for them, rather than monitoring their wellbeing behaviour.

The clearest boundary is simple: employees should feel supported by the technology, not watched by it.

Discover how Wellhub can help yougive employees more choice in how they care for their wellbeing.