Research findings released in March this year, by AI product developers, Studio Graphene, highlighted that 78% of UK businesses claim to be using AI in some capacity. This rises for organisations with - 10-249 employees - to 85%, and only 8% of companies reported not using AI and having no plans to.
The above results, and those from other surveys, combined with lots of news coverage on the topic, mean it’s safe to say, employers are moving fast on AI.
The law, it turns out, is watching closely too.
From hiring algorithms that embed bias into recruitment to performance management tools that miss critical context, the gap between what AI can do and what employers can legally justify is widening, and tribunals are starting to take notice.
We sat down with Charles Hill, Senior Associate at Bird & Bird, to find out where the real risks lie, what defensible AI-assisted decision-making actually looks like, and why the next three to five years could bring a surge of claims that many businesses simply aren't prepared for.
Q: What is currently the biggest employment law risk for businesses adopting AI?
A: The biggest risk for employers using AI to manage their workforce is discriminatory, or otherwise unfair or unreasonable decisions, and outcomes from unchecked automated decision-making. Recruitment tools, for instance, can lead to biased hiring decisions if they are not properly analysed and monitored; similarly, tools used to make performance management decisions may overlook crucial context such as disability-related absence.
The sophistication of AI tools continues to rapidly increase, but the risks remain very tangible and without proper scrutiny of how these tools work and the outputs they produce, employers may unwittingly open themselves up to legal claims and reputational damage.
Q: To what extent can employers legally rely on AI when making decisions about their workforce such as promotion, performance management, disciplinaries, or dismissal?
A: Employers cannot simply hand over workforce decisions to AI. AI can play a useful role in a decision-making process, helping to process and evaluate available information, but an employer's actions must ultimately be explainable and the product of their own reasoning. A decision to dismiss, discipline, or promote someone needs to be justified by reference to the employer's own assessment, not just deferred to an algorithm. Overreliance on AI also risks eroding the mutual trust and confidence between an employer and its workforce.
Q: How can businesses ensure AI systems do not create unlawful discrimination, and what evidence would they need if challenged?
A: There are two key parts to this. First, employers need a rigorous procurement process - looking at how AI tools work, the quality of their training data, and how they address bias. Second, there must be thorough human review of AI output to make sure it is explainable based on the input data, with routine auditing for possible bias. In an employment dispute, AI-generated discrimination will be assessed in the same way as human-generated discrimination, but over-reliance on or unreasonable use of AI may increase the perception that an employer has acted unreasonably and therefore significantly weaken their chances of successfully defending a claim.
Q: What level of human oversight is required to make AI-assisted decisions legally defensible?
A: In our view, decision-making should always be the ultimate responsibility of a human. AI can inform and support the decision-making process, and it is particularly useful in evaluating data to support decision-making, but it should not be making the ultimate call. That means a suitably qualified and informed person needs to review the AI's output, apply their own judgement, and be accountable for the outcome. Without that level of oversight, decisions risk being legally indefensible - particularly in areas like dismissal or disciplinary action, where fairness and reasonableness come under close scrutiny.
Q: What policies and governance structures should every business have in place before allowing widespread employee use of generative AI?
A: At a minimum, employers should have AI use policies in place that set clear guardrails on what is and is not permitted. These should cover which tools can be used, what they can be used for, and how AI-produced output is supervised and monitored. Beyond standalone AI policies, employers should also be updating existing policies to account for AI - for example, setting boundaries for how AI fits into disciplinary and grievance procedures. This is especially important given the huge rise in AI-generated employee complaints, which can significantly increase the burden on internal processes and resource.
Q: What employment claims or tribunal cases involving AI do you expect to emerge over the next three to five years?
A: We expect to see claims coming through in two main areas. First, claims linked to restructuring programmes where the rationale for restructuring is organisational and cost efficiencies driven by AI, or where AI has been used in selecting employees during restructuring exercises. Second, we anticipate a rise in claims around discriminatory and unfair automated decision-making as AI workforce management tools become more widespread - particularly as technologies like agentic AI give AI systems greater autonomy with less human oversight. That being said, AI is increasingly pervasive in the employment relationship and we expect the significant majority of claims to have an AI-related angle in three-to-five years.
Q: What are the most common mistakes you see businesses making when introducing AI into the workplace?
A: The most common mistake is perhaps not having a clear picture of how the business expects its people to use AI. Without that clarity, it is very difficult to encourage productive use of AI while also keeping it within sensible boundaries through policy and training. The result is often inconsistent practices across the organisation, unmanaged risks, and missed opportunities to get the most out of AI in a controlled manner.
Q: If you were advising a board today, what are the three most important steps they should take in the next 12 months to prepare for AI-related employment risks?
A: The three most important steps are: first, audit your existing AI tools for bias so you understand your current exposure. Second, make sure you have the right policies and procedures in place around employee use of AI, including updating existing policies to deal with issues like voluminous AI-generated disciplinary and grievance processes. And thirdly, invest in training - particularly for managers and HR - on decision-making in the context of AI. This will allow the people making workforce decisions to properly understand both the opportunities and the legal risks – including, for example, understanding that the underlying AI prompts used in the decision-making process may be disclosable in employment tribunal proceedings.
If you’re looking for a talented professional to join your team or a new career to help you develop and grow, contact your nearest Reed office today.
