Consultation With Staff and Unions About AI Adoption

AI deployment in HR systems affects employees directly: it determines whether they are hired, how they are rated, whether they are promoted, and in some cases whether they keep their jobs during restructuring. Employment law, data protection law, and good practice require that organisations consult with affected employees and their representatives before deploying AI systems that impact them.

Legal Basis for Consultation

In the UK, there is no absolute legal requirement to consult with employees before deploying AI hiring or monitoring systems. However, there are strong legal incentives to do so.

First, under the Information and Consultation of Employees Regulations 2004, employers with 50 or more employees must inform and consult about decisions affecting employees’ interests, including technological changes. A decision to deploy AI that will determine hiring, performance assessment, or redundancy selection is precisely the kind of decision that triggers the duty to consult.

Second, if the organisation has recognised trade unions, collective bargaining agreements may require consultation about technological change and changes to working conditions. Failure to consult may be a breach of the collective agreement and may trigger industrial action.

Third, from a risk perspective, consulting reduces the risk of legal challenge. An employee who is rejected by an AI system and later sues for discrimination may argue that the organisation was negligent in deploying the system without testing for bias or consulting about its impact.

An organisation that can point to consultation, testing, and employee feedback is in a much stronger position to defend the decision.

Effective Consultation

Consultation is not a tick-box exercise. It means genuinely seeking employee views and being willing to modify plans based on feedback.

Effective consultation involves:

describing what system the organisation plans to deploy and why (what business need does it address), explaining how the system will work, explaining what data will be used, explaining what decisions it will make or influence, explaining what risks it poses, inviting employees to raise concerns, listening to concerns and taking them seriously, and explaining what changes will be made based on feedback.

Consultation should happen early, before the system is purchased. Once a vendor is selected and a contract is signed, it is too late to consult; the organisation has already committed to the system.

Consultation should involve representatives of affected employees. If the system affects hiring, recruitment teams should be consulted. If it affects performance management, line managers and employees should be consulted.

If it affects monitoring, employees across the organisation should be consulted. If unions are recognised, union representatives should be part of the consultation.

Common Employee Concerns

Employees typically have several concerns about AI in HR systems. They worry about fairness: will the AI treat them the same as others? They worry about transparency: will they know they have been subject to an automated decision?

They worry about their ability to challenge decisions: can they ask a human to review if they disagree?

They worry about monitoring: will the organisation use AI to spy on them? They worry about job loss: will AI replace them? They worry about data: who has access to their personal information?

These are all legitimate concerns, and an organisation that listens to them can often design safer, more compliant systems. For example, if employees express concern about transparency, the organisation can commit to telling employees when an AI system influences a decision affecting them. If employees worry about monitoring, the organisation can commit to limiting what is monitored and providing employees with access to their monitoring data.

Using Feedback to Improve Systems

Consultation is only valuable if the organisation acts on feedback. If employees raise concerns that have not been tested for bias, the organisation should commit to bias testing before deployment. If employees worry that an AI system will disadvantage people returning from parental leave, the organisation should specifically test for adverse impact on employees with employment gaps.

If a significant majority of employees oppose deploying an AI system, the organisation should seriously consider whether to go ahead. Deploying a system despite widespread employee opposition is likely to damage trust and may trigger legal challenge.

Ongoing Engagement

Consultation should not be a one-time event. The organisation should maintain ongoing dialogue with employees and union representatives about how the AI system is performing, what problems have emerged, and what changes might be needed.

This might include quarterly updates to all staff on how the system is performing, annual review of monitoring data to check for emerging bias, and an open process for employees to raise concerns about the system.