The rise of AI-generated claims in employment tribunals

In this episode of the Employment Law Pod, employment Partner Andrew Whiteaker is joined by dispute resolution Partner Phillip Baldwin to explore the growing impact of AI on litigation, grievances and dispute resolution.

Andy and Phil explore how litigants, employees and businesses are increasingly using tools such as ChatGPT to prepare grievances, legal arguments and tribunal claims. They discuss how AI can improve access to justice and support individuals who may otherwise be unable to obtain legal advice, while also creating significant challenges for courts, tribunals and employers. 

The conversation covers:

• Why employment judges are starting to push back on AI-generated submissions
• The rise in lengthy and overly complex tribunal claims
• How AI can create false confidence for litigants
• The risks of hallucinated case law and inaccurate legal arguments
• Confidentiality and privilege concerns when using open source AI tools
• The growing pressure on tribunal and court systems
• Why AI can make settlement discussions more difficult
• Practical advice for employers handling AI-generated grievances and complaints
• The continued importance of legal expertise, negotiation and human judgment in dispute resolution

Alongside the risks, the episode also looks at the positive ways legal professionals are using AI to improve efficiency, including disclosure exercises and document review, while emphasising the importance of human judgment, negotiation skills and legal expertise in resolving disputes effectively.

If you have any questions about employment disputes, grievances, litigation or the use of AI in the workplace, the teams at Boyes Turner would be happy to help.

Episode Links

• Visit our website
• Contact us
• Our Employment services
• Our HR training

Transcript

Andrew Whiteaker 0:03

Hello and welcome to the Employment Law Pod from Boys Turner. My name's Andy Whittaker. I'm a partner in the employment team here at BT. And as ever on the pod, we like to talk about things that are interesting, maybe some new cases that we've seen, some legislation that's been passed, or some interesting developments in the field of employment law. And today I'm pleased to say that I'm joined by a guest, not from our employment team, but today I have with me a Phil Baldwin, a partner from our Dispute Resolution team. Hi, Phil.

Philip Baldwin 0:32

Hi Andy, how are you doing?

A Judge Pushes Back On AI

Andrew Whiteaker 0:34

Yeah, very well, thanks. Very well. And the reason that I have Phil joining the podcast today is that we are both litigators in our own fields. And we have both experienced over the last year the increasing impact of AI on our workloads, both in terms of how it's used by litigants in the tribunal and the civil courts, but also in how individuals are seeking, or businesses even are seeking, to use AI in a pre-litigation stage in respect of the letters before action or their grievances, whatever it might be, and how that is driving some complications and challenges for our clients. So I think maybe if I can give a bit of context of the sort of things that I'm seeing, and then Phil, maybe if you sort of give a few examples that you're seeing too. And what really triggered me to have this conversation with you today, Phil, is some correspondence that I received from the tribunal in a matter that I'm dealing with just last week, where I'm acting for one of the respondents in the proceedings, there is another respondent, and there is a claimant, and neither the other respondent nor the claimant have any legal representation. And as a result of that, it was quite clear to me previously, and it is also now obviously dawned upon the tribunal, that both of those parties are leaning heavily on the use of AI in the way that they're conducting the litigation. And this resulted in a the first time I've seen a letter of this nature from the employment judge to the other parties, this was following a direction that had been made for the parties to make comments on a list of issues that have been prepared and spend no more than two sides of A4 doing it. The tribunal wrote to the party saying, in turns, I'm paraphrasing slightly here, it's very obvious to me that both parties are using AI to support them in this litigation. Clearly, AI can be a useful tool in gaining knowledge and understanding, but your use of it is resulting in overly long, overly complex, inaccurate correspondence that's not actually addressing the points that I'm asking you to address. In particular, I ask you to spend just two sides responding to this list of issues. The response that I received was 10 pages long, didn't respond to the list of issues at all, and just set out your case again. If you're happy with a list of issues, just reply to this email saying, I am happy. And please, please, please stop using AI in this fashion because it is driving up costs unnecessarily and it is causing matters to be much more complicated than they need to be. I I don't know if you've had a sort of similar experience at all, Phil.

Philip Baldwin 3:18

Yeah, so what an interesting 12 months it's been. What we see quite a lot now. And obviously we're in different forums, Andy, are we? In terms of you and you being in the tribunal, myself being in the in the courts, civil courts, high courts, county courts, etc. Well, what we have seen a lot of is my initial instruction, it's not hello, can I have a chat with you? It's here is my legal case, here are the strengths, here are the weaknesses, this is what I think we should do. This is the leading case law. Can we now have a chat? So, what what we get straight away is is something from AI. And I think it's persuasive AI to a certain extent. It often sounds very authoritative, and they are they become very confident. And one of my major issues at the moment is managing expectations of of clients when using AI. I'm sure we've all put things into chat GPT, and we are we are we are rarely wrong when we ask Chat GPT. It normally says, Andrew, well, thank you for those comments. That's a great idea. Could I perhaps recommend X, Y, or Z? So we've seen a lot of that, and it's it's that question, and we could probably talk about this for a couple of hours, I'll be honest. But I think it's that question is that does AI genuinely empower litigants or does it create that sense of of competence, which unfortunately they may not actually have? And I think that's that that's the question at the moment is that are we giving people better access to justice? And I know that you've had an influx of tribunal claims off the back of the AI generation. Is it is it really giving litigants in person access to justice? I think there's arguments for for both sides of that coin.

Andrew Whiteaker 5:02

Yeah, I think that's right. I mean, what what we can't be obviously is a couple of old duffers howling at the moon here saying, you know, get rid of AI, boo-boo, it's bad, because there's frankly nothing that you and I can say that's going to change anything in terms of the increasing use of AI by litigants in in the in the court system. And as you quite rightly point out, if I use a scenario maybe in the employment field where you have someone who has lost their job, they don't have lots of savings, they don't have an ability to obtain legal advice through the normal means by paying for a solicitor to support them, with the tribunals being in such a state as they are at the moment with extraordinarily long delays in getting to any hearing. If you're that individual who otherwise can't obtain legal advice and can't get access to justice, then who are you or I, or anyone for that matter, to be there saying you shouldn't use this as a tool to provide you with confidence, to provide you with the knowledge, to provide you with guidance as to the sort of claims that you should pursue. So that there is definitely a place for it. But as with you, you know, I am receiving queries from clients where they are saying, well, I've just run this through AI and this is the answer that I've got. Can I just double check that with you? Now, internally at Boys Turner, we have access to some AI tools, and I find them very useful. I do find you know it helpful to run a query or to run a document through it, just to have a quick initial set of electronic eyes on it to highlight things. But I mean, this this broadens the discussion out a bit more a bit more broadly, I suppose, in with in respect of careers and jobs and professions, etc. But you and I, with the experience that we have, when we see something produced by AI, we can know what is maybe a hallucination or maybe something that's a product of an inaccurate prompt and not the not the the right question not necessarily being asked or being asked in a slightly different way, or if there are issues that have been raised that are unrelated or irrelevant to the particular matter, and we can weed through that because we have the experience and the expertise in dealing with matters of this nature. But of course, if you don't have that training and you ask a question that's not quite on point, or that the prompt is not necessarily 100% accurate, you get, as you've said, Phil, very authoritative advice through, well, this is the answer, this is the correct answer. And if you're in receipt of that advice, then why would you listen to anyone else who tells you that you're wrong?

Access To Justice Versus False Confidence

Philip Baldwin 7:41

And this is the point in terms of sort of alternative dispute resolution and trying to manage the client's expectations. It's the times have changed slightly because people come to me slightly more aggressive now because they've been told by persuasive AI that they've got a very good case. And if there is a bit of a bump in the road in their case, they will use AI and then tell me how to deal with it. So I think it's just navigating that and going back to sort of hallucinations and fake cases and things like that. There has been quite a lot of articles and cases of uh unfortunately barristers, believe it or not, being being caught using, not caught, I think, perhaps not validating or confirming those authorities that they have used. And they they've definitely got in trouble with their with their respective regulators. Now, obviously, from the from the civil court perspective, there is no at this current stage no mandatory requirement to declare the use of AI. So I don't need to go into a judge and say, oh, just to let you know, judge, I I've I've used used AI on this. I think that might change. And I think there are consultations and proposals going on as to whether a declaration is required. But what it goes back to is our qualifications, Andy. It goes back to we are on the hook if there is a mistake when AI has generated something in respect to the legal case. And I think it will remain that way because I can't see how else it can change. But in terms of a litigant in person, that's that's slightly different, isn't it? Because obviously we can be held accountable if we if we make mistakes. Where a litigant in person, it can be far more difficult to sort of unravel things if the litigante person is using AI so heavily. Some of the good ones, Andy, that I've seen very recently is a witness statement which which obviously shouldn't have case law or for authorities in, and that regularly happens when using AI. I also have quite a lot of clients that will look me straight in the eyes and say they haven't used AI, but certain headings are bold, there are some long hyphens in there, there is not one spelling mistake to be to be seen. So I think there is a bit of oh no, I I haven't used AI when they have in fact used AI. So I say it there's some really, really good ones at the moment, and we've had uh things from injunctions where they have just used AI and in fact still got the brackets of enter the appropriate name here. So we we've seen we've seen a lot, and as I say, I think it is positive. I think it gives people a good basic understanding of of what is required for certain procedures, etc. But then we come down to the consequences when when a mistake is made. Now now there are real consequences, and as I say, I believe some barristers have been struck off as those consequences. Now, what are the consequences for litigants in person? Well, there are cost consequences in the forums I operate in, and you can be heavily punished if you uh if you lose a case or don't deal with it properly or procedurally. So there are cost consequences in using AI, which I I don't think sometimes litigants in person are fully aware of.

Andrew Whiteaker 10:55

Yeah, and and one of the challenges that we have in the employment tribunals at the moment are the the lack of those consequential cost consequences, if you see what I mean. It is the case that costs can be awarded against a party where they've acted vexatiously or unreasonably in the conduct of proceedings. But what we are seeing happening in respect of the expanded use of AI in litigation is first of all a massive increase in it. So we've seen an almost doubling in the caseload before the tribunals since the pandemic. Now, we know that the Employment Rights Act is being introduced and there's new rights that are coming on stream over the next year or so, but this is all in advance of that. So this is not ERA related, this is caused by something else, and the best guess guesses rather are that it is AI generated. And it's not just the amount of claims that are being pursued, but it's their complexity. And again, you know, you you ask uh an open source model, I've been, you know, my boss doesn't like me and he hates me, and they've treated me this way because I'm this is my characteristic, etc. etc. And then you get told that you've got cases for every form of discrimination under the sun, and then you ask the uh bot to help you draft pleadings, all of a sudden you have extraordinarily long pleadings that are claiming every possible form of discrimination that you can imagine. And I don't say this to sort of suggest that all claims pursued in this way are unmeritorious, but especially if you're looking at claims of discrimination, unless they are clearly misconceived, tribunals won't strike them out at an early stage because they'll say, Well, we need to hear from the parties in evidence to evaluate the the value of that evidence and to take a view as to who may be telling the truth and and what the veracity of their statements might be. So you get these extraordinarily complicated claims with numerous heads involved, and although you and I might look at them and say, Well, I just these these aren't strong claims, I don't think these are necessarily likely to succeed, they're nonetheless having to be dealt with in the tribunal process with a limited scope for them to be struck out, even if they're not the strongest claims. So that's causing a huge backlog in the tribunal system. We're seeing if you're trying to issue a claim in London South, for example, and you need a five-day hearing, the latest stats are that you won't get that until 2029 now. We've even seen cases being listed for 2030. So this is causing a huge pressure to the tribunals. And to go back to the original point, I suppose, around costs is that even if you are unsuccessful, the default position is that you bear your own costs. So you normally do not have to pay for the other side's costs unless you've acted uh vexatiously and unreasonable. And bringing a claim that's not the strongest, but is arguable, is is not going to cross that threshold.

Hallucinations Accountability And Cost Risks

Philip Baldwin 13:50

Yeah, we discussed that before, happening, Andy, the the lack of teeth in terms of in terms of costs in the tribunal. And obviously a slightly different in the in the in the civil courts. Now, well, one of the points that you touched on there is that they will they will plug this into an open source AI model. There must be issues, and again, I don't think litigants in in person consider the the confidentiality side of things in terms of these these open source AI models, and and what what we get and what will likely happen is that full witness statements, full defences, case summaries will all be plugged into AI tools. And the litigant in person, most likely, has no idea or understanding where that information goes. There's legal professional privilege issues as well, which again can be a completely separate podcast if we wanted it to be. But those are all things that I don't think litigants in person consider. You can't just go typing in confidential information that perhaps they'd been privy to during their employment into open source AI. That could potentially lead to claims against them further down the line for some sort of breach of confidentiality and some losses that flow from that. So again, I think that's an area that we will see more and more touched upon as sort of the law, not the law, but I think perhaps the understanding of how AI is going to be used in both your forum and mine.

Andrew Whiteaker 15:19

And I think that segues nicely, I suppose, away from you and me telling war stories about how it's impacting us and maybe giving some some practical advice about what businesses can be doing to best manage how they're running up against AI, either in managing their employees or in respect of complaints or or commercial disagreements, I suppose. And and dealing with with my forum first, where we run into it most frequently to begin with is in the form of grievances. So it's where someone is unhappy about the way they've been treated and therefore they rely upon AI to produce a grievance for them. And the reality is that these grievances that are received are often long and repetitive, they can be overly complex and legalistic in tone, include those exaggerations that we spoke about as well. And you know, having read them myself when my clients received them, you it's really difficult sometimes to actually get to the bottom of what the individual is actually unhappy about because there's just so much content in there and case law being referred to and and statute, and you think, well, but what is what what's the actual issue here? So, what can businesses do to try and address this? Well, I think the first thing you can try and do is perhaps just try and avoid a grievances to begin with.

Philip Baldwin 16:38

Yes, it's always helpful and and all and all claims as well from from my perspective.

AI Driven Claims And Tribunal Backlogs

Andrew Whiteaker 16:43

Absolutely. So making sure that our managers are well trained in in soft skills and in managing their their teams is clearly an important thing. Making sure that we have an open culture where individuals feel able to raise concerns at an early stage and seek resolution to them rather than things being allowed to fester and resulting in formal grievances. We deal with those complaints as and when they may arise. And also encourage the use of informal resolution rather than it escalating to a to a formal grievance. Then I think when you do end up having to deal with a grievance, my recommendations really be don't let the individual hide behind AI. So we were actually seeing an increase in individuals saying, Well, I can't come to a grievance meeting, but here's my here's my grievance. And then you try and get them to come in, they won't. You produce a response, and then you get an appeal, which again is very clearly generated by AI, but they then won't come to a meeting. Try and have face-to-face discussions because actually that will help you really get to the bottom of what the actual issues are. It might be, and this might be being me being hugely unfair, but it might be that individuals have got content produced for them by AI that they don't necessarily understand fully, and that may be running arguments that they don't actually intend to run, or not putting it in exactly the way that they would put it, but relying upon that content that's being produced. So actually having that opportunity and focusing on meetings to speak to the individual, not as a way of intimidating or anything like that, but just to try and understand in their own words what the actual issues are rather than investigating a word salad that's produced via AI from time to time. Obviously, we need to continue to focus, follow our codes of practice, our internal policies, ACAS codes, that sort of thing. But then when we're responding to complaints that are made and grievances that are that are raised, we don't necessarily need to do a point-by-point rebuttal. Again, you know, if people are running legal arguments or quoting cases, this is not the forum for that. Okay. You're upset about something, I'm looking into it. I don't need to produce a you know 10-page legal rebuttal of the issues that you've raised and relying upon case law. I'm just trying to understand what you say went wrong and whether we agree with that or not. So don't get dragged into that. But do take things seriously. And I've seen this over my years of practice that sometimes where clients fall down is when you have a, and I'm using the phrase very loosely here, but a known troublemaker or an individual who's seen as problematic or difficult within the workplace, and they raise a grievance and you give it short shrift because you're annoyed with them, oh not this person again. And you don't necessarily engage with the relevant and reasonable points they're making just because you're annoyed with that individual for other transgressions or the way that they might have conducted themselves in the past. So just because it might be an AI-generated grievance that is overly long, massively exaggerating, raising points that are uh irrelevant and that are repetitive, that doesn't mean that there's nothing in it. And that doesn't mean that you don't have an obligation to address it and carry out an investigation. And the final point I'd make is just around the issue around confidentiality that you raised. Yeah, in our internal policies and procedures, whether it's around the use of confidential information, whether we have an AI policy as well, making it clear to individuals that there are limits to how they are allowed to use those tools in work or elsewhere, and also in the use of confidential information and what is and is not prohibited to avoid those scenarios, like you say, where someone wishes to raise a grievance and then ends up putting client confidential information or data protection issues as well, isn't there? Yeah, information in respective third parties into an open source AI model, and then that just being disseminated wherever it might end up. So I think those are the sort of things from an employer's perspective, dealing with tribunal matters, what what they should be thinking about. I mean, from your perspective, have you got any other suggestions that you would make or to your clients about how best managing these are.

Confidentiality And Privilege Pitfalls

Philip Baldwin 20:59

Yeah, so so obviously and my my clients are more corporate commercial clients. They have disputes that come up, and sometimes I do see directors or or board members try to deal with those disputes and perhaps they lean on AI. And I think what I would say is that it's still beneficial, and I think will always be beneficial, that responses, even if they come from the company in the first instance, are run past uh uh solicitor. And I know we do this quite often where we are in the background in the early stages in order to make sure that some of the arguments that they are running are correct, to make sure that documents and letters aren't peppered with uh authorities and statutes where where we're trying to, especially for our commercial clients, we're trying to deal with things on a on a commercial basis. Uh we don't want to get tied up in lengthy litigation. So I think that my my first point is that we are still very relevant despite the despite the the growth in AI. I am a firm believer that that the solicitors, lawyers, barristers, etc., will be required from a dispute resolution perspective for many, many years to come. Also, with that said, is that I haven't seen AI very good at settling cases.

Andrew Whiteaker 22:12

No.

Practical Employer Guidance On Grievances

Philip Baldwin 22:13

He's very, very good at uh saying how good your case is and perhaps how bad the other side's case is. But in terms of being able to bring parties together, perhaps getting them around a table or on a team's call or a mediation of some sort, unfortunately, our friend ChatGPT can't do that. And I think there is a a lot of value in how we're able to do that, how we're able to try and bring the parties closer together. Because ultimately a lot of my clients don't want that long, protracted litigation and the costs that go with it. They are happy to try and conclude matters quicker. And I think that is where that added value is. So, yeah, from my perspective, is that I think there is still quite a lot of value in getting in touch with your solicitor or or a solicitor at an early stage, especially when there is a dispute rumbling on. Unfortunately, just plugging it into an open source AI isn't going to give you what you need to settle that dispute. There's also the court element to it as well. At some point, you perhaps have to sit in front of a judge that will look at these documents. And if you don't know that you have accidentally put some Australian or New Zealand case law into a document, then of course you will look silly. And if the other side are legally represented, they may be looking to recover their costs from you, whether you're a litigant in person, a company, or whatever. So there are real life consequences to making mistakes. I don't think AI is at a position now where mistakes aren't going to be made. I see them quite regularly. I think there is a use for it. I've just gone through quite a large e-disclosure exercise where we've we've harnessed AI in order to assist with that disclosure exercise rather than having many, many of the team sort of locked away in rooms doing the disclosure process. We've been able to use AI in order to become more efficient. But unfortunately, the good old-fashioned getting somebody around a table doing a deal and and bringing uh a matter to a conclusion quickly, I think is still very much reserved for people like us, Anne.

Andrew Whiteaker 24:20

I I agree. It's one of the challenges we really face in terms of settling lots of our tribunal claims at the moment, where individuals have been told, well, you will win and you will win this amount of money.

Philip Baldwin 24:29

Yes, it's so difficult.

Andrew Whiteaker 24:31

And you know, how how do you argue with that? Because that's what AI has told me, and that's therefore what I'm going to stick to. And you know, I know the the rules, although the CPR does exist in the tribunals, we have a degree of flexibility in the way that they're interpreted in a way that is not always available to you. But, you know, for example, if if you're not going to get your witness statements over the line in time, it's not impossible to drop an email to the other side and say, Well, I'm not going to get these done by Friday, but I can get them to you on Wednesday. And the solicitor on the other side says, Oh, yeah, no, it's fine, don't worry, because they might have to ask a similar question of you in respect to another direction, and they know that actually the hassle and the cost of complaining to the tribunal about what's by the time a decision is made that documents have already been exchanged. But when that's say in that same matter, the individual is not legally represented and they are ask an AI chatbot about it, and they're told no, you should say no, and you should write to the tribunal about how terrible this is, then that's not necessarily assisting the parties move things forward. It's that experience and that knowledge of how things actually work in practice that I think is is still a really valuable, valuable asset that can be brought to these sort of processes.

Philip Baldwin 25:43

Yeah, and look, I think I think we're going to see a lot of change and we're going to have to change the way we practice, the way that we operate. But yes, it's it's been a it's been an interesting time. And as I say, there is very much a place for AI. And I know that we use it through our formal sort of legal AI systems that that we have at Boyce Turner, but the use of open source AI and your your your co-pilots, your chat GPTs, it's going to be interesting to see not only how we deal with it, but how the courts deal with it further down the line. I think we will get to the stage where there will be you'll have to provide some declaration that you have used AI and how you've used it. I know there are some consultations going around at the moment with that in mind. That seems a very, very sensible approach. And again, it will be very interesting to see whether litigant and person will be held accountable for mistakes in in AI-generated documentation. When when there is a litigant person on the other side, the courts have to be quite uh not sympathetic, but but helpful to the litigant and person. Perhaps that may change in the future if we've if we've got AI. So, yes, lots to talk about, and probably we should catch up in six months' time, Andy, and see where we've got to on AI generally.

What Changes Next And Closing

Andrew Whiteaker 26:56

Yeah, that sounds like a plan. So that probably does wrap us up for today, Phil. Thanks. Thanks so much for joining us on the pod.

Philip Baldwin 27:02

Thanks for having me.

Andrew Whiteaker 27:03

No problem at all. And thanks also to everyone for listening to this episode of the Employment Law Pod. If you're interested in checking out any other episodes, they can be found on the Boys Turner website. Or alternatively, you can follow or subscribe wherever you might pick up your podcasts. Thanks again, everyone, for listening, and see you all again soon.