The costly illusion of a convincing AI answer in employment law: Simon Robinson
For a busy employer facing a tribunal claim, a chatbot offers an instant answer, neatly structured and confidently expressed. That apparent certainty is seductive, but it is also where expensive mistakes begin.
Tribunals are already dealing with the fallout. AI-generated claims and submissions can be lengthy, repetitive, and difficult to follow. One skeleton argument submitted to the Employment Appeal Tribunal attracted criticism after running to 300 pages. The problem was not simply excess detail; it was that the volume obscured the issues, increased costs and the claimant expected the other side to spend checking the accuracy and legitimacy of the content. Rightly, the Employment Appeal Tribunal roundly rejected the appropriateness of such an approach. In June, the High Court issued a warning after dozens of false cases were cited.
Some authorities did not exist, while others existed but invented passages from them were quoted. A document can look professional while resting on fiction and an employer who files it owns the consequences. Tribunals now regularly enquire as to the extent to which parties have used AI in preparing their cases.
Employment law is not about producing a superficially impressive document. It is about making the right call: which points to fight, which to concede (on the basis it does more harm than good to contest a point you are inevitably going to lose) and when to settle. I call that experienced judgement or my alter ego, “grAy I.” GrAy I is 54 years old has 27 years’ experience in employment law and it can tell a client, “Yes, you can do that, but you shouldn’t.”
A specialist evaluates the evidence, prepares witnesses, and assesses the commercial reality. Sometimes a witness meeting changes the whole strategy. Hesitation, inconsistency, or a missing document may point towards settlement. Software can process the information it receives, but it cannot reliably spot the significance of what a person avoids saying.
A chatbot also accepts the assumptions built into the question. Ask it to defend a dismissal and it may oblige without first asking whether the process was fair, the evidence reliable or the manager consistent. Good advisers challenge their clients.
We look for the awkward fact, the absent email and the alternative explanation before the other side does. People may also input into a chatbot the case they wish they had rather than the one they actually have which leads AI giving advice and submissions on cases that don’t actually exist.
AI still has a useful place. It can draft a job description, summarise a policy, check spelling or help organise an initial chronology. The mistake is to confuse assistance with assurance. Every output needs checking, and every important decision needs someone prepared to take responsibility for it. Legal budgets are tight and the attraction of a free answer is obvious, but the old adage applies – if you think hiring a professional is expensive, wait until you hire an amateur. AI for all its abilities is very much an amateur. Yet the cheapest document can become the costliest problem if it misses a deadline, weakens a defence, or closes the door on an early resolution. The sensible question is not whether AI can write it, it is whether anyone with experience has challenged it.
By Simon Robinson, Employment Solicitor & Director of Robinson Ralph