UK business owners are being advised on the potential risks of "vibe lawyering" – using AI tools to build legal arguments without formal training or support – as growing numbers of employees and employers are turning to AI to "lawyer up" in workplace grievances and complaints.
The warning comes as employment tribunal claims have reportedly increased 39 per cent year-on-year – a rise specialists partly attribute to the growing use of AI in workplace claims as the technology makes it faster, cheaper and less intimidating for employees to act on a grievance than ever before.
SMEs are particularly at risk because, unlike larger organisations, they rarely have in-house legal or HR expertise to fall back on, making them more likely to turn to free or low-cost AI tools when navigating complex employee issues without the necessary expertise to contextualise the results, or determine where AI might be wrong.
Gill McAteer, director of Employment Law at Citation, says the fundamental problem with using AI and LLMs like ChatGPT in workplace disputes in particular is that employment law relies heavily on fair judgement and reasonableness being applied to each case's unique circumstances – qualities AI cannot replicate.
"Where AI really shines is speed – summarising a mountain of correspondence or giving someone a sensible starting draft in minutes rather than hours,” she says. “That's a real win for time-poor SMEs. But employment law is different because so much of it is underpinned by nuance and the concept of reasonableness, which AI still struggles to grasp.
“Although they’re becoming increasingly sophisticated and can be brilliant in certain use cases, from our own tests we know these tools still get legal answers wrong in a considerable proportion of cases and because the answers are so convincing, mistakes can be very hard to spot without expert knowledge.
"As employment law professionals, we spend a lot of our time ensuring we establish the correct context for a matter, like the specifics of the employee and the relevant background information. People rarely know what information is important, so LLM’s fail to provide correct guidance as each case is unique.”
But Gill says this is increasingly playing out on both sides of workplace disputes. Employees are using AI to build a case from a one-sided account of events, then becoming emboldened when they are perhaps wrongly told they are entitled to compensation or action that isn't supported in actual law.
Whereas grievance letters written by employees tend to be a couple of pages long, AI-generated grievance letters are now stretching to fifteen or twenty pages and are often so repetitive and bogged down in technical terminology that employers "can't see the wood for the trees."
Business leaders are also at risk when they use AI to draft responses to complaints or dismissal letters without a qualified person checking the content first.
Citation's experts warn that AI-generated correspondence can actually be used against employers at tribunal, if wording suggested by AI hands the other side legal ammunition.
Gillian McAteer Photo: Handout
Gill cites probationary dismissals as a typical blind spot. She said: “AI will readily and accurately walk an employer through a standard dismissal process step by step – that part it does well. What it has no way of flagging is that the same employee raised a safeguarding concern weeks earlier.
“If this fact is overlooked by the employer inputting the information, it can quickly turn a routine dismissal into a potential whistleblowing claim worth tens of thousands of pounds in damages."
To help businesses owners who are considering using AI on workplace issues, Citation has provided the following advice:
1. Treat AI-generated answers as a starting point, never a final position
AI can be useful for a quick first draft or basic information, but it doesn't know your business, your policies, or the specific circumstances of your case. AI also tends to be unreliable when it comes to understanding recent legal developments, so don’t rely on it for this information, particularly at a time of continuous change with the ERA (Employment Rights Act) reforms.
2. Never send AI-drafted correspondence to an employee without it being reviewed by a qualified human
AI-written letters can read well but often bury damaging wording, ramble past the point, or jump ahead of where the process actually is – for example, raising the prospect of dismissal when it's not been discussed as an option. Once correspondence has gone out in the business's name, it can then be used against the business at tribunal.
3. Be alert to one-sided, AI-generated grievances or claims, and ask the questions AI won't
AI takes what it's told at face value and builds a case from one person's account. Regardless of how confident a complaint may come across in terms of entitlement or outcomes, remember to carry out your own investigations and consider all sides.
4. Remember AI can't weigh up individual circumstances or wider background – getting this wrong can be costly
Two cases that look identical on the surface can call for very different responses depending on factors not obvious at first glance, such as concerns raised by an individual or issues regarding their health. Missing that context can turn a routine decision into a claim worth tens of thousands of pounds in damages.
5. Use AI for speed on simple, factual questions, but bring in a human for judgement and strategy
AI can help with summarising documents or pulling together background, but employment law relies on judgement and reasonableness applied case by case, where AI falls short. So, use it for the groundwork, but bring in a qualified person for any decision that needs to hold up under intense legal scrutiny. Decision-making itself must always lie with the employer and not be delegated to AI.