This is the Mills & Reeve employment law update podcast.
I'm Michelle Clarke.
And I'm Richard Barker.
And in our last episode, Jameel and Neil shared their experiences from the front lines of dealing with investigations. If you haven't already listened, please do go and give it a listen. In this episode, we're going to look more closely at what is seemingly a growing trend of employees' use of or reliance on AI when identifying and writing complaints and other documents, and also the potential pitfalls and additional considerations that employers and, in particular, HR teams should be keeping in mind, and our tips for navigating those uncertainties.
So, Richard, I guess as a starting point, what are your experiences of seeing the use of AI crop up in things like grievances? Yeah. I think probably six months or so ago, certain clients thought, "Is it just us that are getting these grievances that have clearly been drafted by a robot?" And it's not, and it definitely isn't now, and actually I would say it's probably quite rare now that where we're advising a client that we see a grievance that probably hasn't been put through some form of AI, whether it's Copilot or ChatGPT.
Think that all that is ultimately doing is increasing the complexity and increasing the amount of time that needs to be spent on dealing with that because the very nature of AI is it provides you with further prompts, further questions, etc. So I guess in terms of what we're seeing a lot of it, we're seeing it across the board, and we are not really seeing what perhaps you did see in the past, which was a short two or three paragraph grievance letter saying, "This is my problem. This is what I'd like the outcome to be," which is quite easy to deal with really from an HR perspective. Yeah, it does certainly feel that the day of a one-page grievance has gone, and everything seems to be at least the 10 pages and the many numbered paragraphs. And also, the use of a lot of case law seems to creep in as well.
Whether or not it's correct or it exists, it does seem to creep in a lot more in a grievance that there's going to be a case or a bit of legislation that's mentioned. And I suppose from the perspective of our clients and the HR teams, it's getting their head around how do we actually deal with a grievance that's quite clearly being drafted by AI.
And in terms of a starting point for clients who are faced with an AI-written grievance, what tends to be your top tip for a starting point on how to approach dealing with that grievance? Yeah, I think that it probably remains what it used to be pre-AI, which is we've got to get to the nub of what the actual issue here is.
But I think the difficulty is how AI almost muddies the waters in that respect. And so if I just give you... I've done a little example where I've used Copilot myself, and all I've put into... I'm not going to read every single part of this, but all I've put into Copilot is, "I am a 45-year-old male, and a 36-year-old female colleague has got a promotion that I felt I should have got. Can you prepare me a rough draft grievance letter?" So no real substance, just named two protected characteristics and moaned that I didn't get a promotion. And then it has prepared a draft grievance letter, but then which it breaks down a lack of process and transparency, fairness, concerns about the selection process, lack of feedback and transparency, impact on my position and development and requested outcomes, none of which I put in the prompt.
It then said, "Can you create a more assertive version?" If I'm an employee wanting to make my point, I'm never going to say, "No, thank you." I'm going to click, "Yes, please." So then it makes it more assertive. And then as we go through the chain, then it says, "Would you like me to make it more legally precise?" So of course I say yes.
So then we start getting the ACAS Code of Practice being brought in.
And again, all the different subsections. It then says, "If you want, I can refine this further into a version that subtly signals potential discrimination concerns without explicitly alleging them." So again, I've said, "Yes, please." So then it's done that.
And I'm not going to get on any further with this, but it says, "If you want, I can now refine this into a version that more strongly hints at a potential indirect discrimination without making an allegation." And you can see through that how it just flows and flows, and how someone who has an issue that they can't really particularly articulate other than that person's got something I wanted, suddenly becomes a grievance letter which is alleging sex discrimination and age discrimination.
So I guess that is the issue that employers and HR teams are grappling with, that actually suddenly you have an employee who has a grievance letter that doesn't even represent what they were initially thinking. Yeah. And that's why I think it's really important that HR teams do as part of any kind of initial investigation into the grievance, sit down with the employee and say, "Look, I know we've got your 10 or 15 or 20-page grievance letter.
Can we just understand in your words what the issues are?" And I guess the difficulty is that we're increasingly hearing our clients say that when you get into that scenario, the employee can't actually do that. They can't actually articulate or at least follow up on what they've created in writing. And I suppose there's therefore a number of options that HR teams can take, but we're living in a fairly new world where lots of this hasn't been tested at tribunal yet in terms of the approaches that people take.
And so it is... it's new for everyone really. Yeah. And there's the reasonable element, isn't there, almost? As you say, if we... if something goes worst case scenario, we end up in an employment tribunal, and we've taken all the reasonable steps we can to try to identify those terms of reference, which I know Jameel and Neil touched on last time in terms of the importance of the terms of reference and doing your investigation against them.
And if you're faced with a document and an employee who, as you say, has tried really hard to articulate what the grievance is, but might not know how to put that in there, AI has then generated discrimination. Whistleblowing comes up quite a lot, I think now. Yeah. And it often gets put into that box, and perhaps the employee doesn't really understand the concepts of them.
But I think there's that element as well, if you're not sure, and it's completely understandable, if AI generation is telling you, "I think you've got these complaints," and then your employer is trying to identify a bit more clearly what the complaints are, you might feel almost a little bit, "I'm going to stick with what AI has told me is going to be those points," so you cannot change anything at all.
In which point you've then got a grievance consisting of 50 different elements, whereas the example you gave, it started off as one, that I didn't get promotion, somebody else did. And it's difficult to navigate, isn't it? I guess almost if you try and take those steps, but the employee goes against them, it's almost what do you then to try and get that process over the line? Yeah, exactly that. And I think sometimes you do have to go back to basics and the ACAS code around disciplinaries and grievances, it does talk about reasonableness, but that has to be from both sides. It talks about reasonableness from the employee perspective as well. And look, there's probably plenty of employees who have a genuine grievance, and they have used AI appropriately to be able to express that grievance in slightly clearer terms, and that's great if it helps in that respect.
I guess the difficulty is that there's plenty of employees who are not using it in that respect as well and are ultimately getting to a point where there's probably elements of their grievance which are unreasonable because it's no longer actually what they believe or what they reasonably believe, to use the legal language, and they can't back it up.
And I think that is what employers are going to be grappling with, which is how do we deal with that? We have a set of legislation which says if someone raises a grievance, we have a process that we need to follow in dealing with that, but is that process fit for purpose in this respect?
And there's nothing wrong with sitting down with the employee and talking through that, but I do wonder whether we might get to a point where some employers are saying, "Look, unless you are able to distil your 20-page grievance down to three or four key points," use the language of terms of reference, however you want to do it, but three or four key bullet points, "we're actually not going to be in a position where we can go and look at this because it's not reasonable to expect us to do so because you're not able, once we've sat down with you, to even articulate this.
So we don't believe that this belief is reasonably held, and we're not going to take it any further." It might be seen as brave, but it might be seen that actually There's degree that any kind of tribunal will have a degree of sympathy with that approach because they're feeling it as well.
We're seeing it within the tribunal system as well. And yeah, definitely. A lot of times, you're now seeing the claim form, whereas before it might have just been completing Section, I think, out of my head, 8.2 of the claim form, and you'd get a few paragraphs in there. Which in itself can have difficulty because then down the line, you're having to clarify what the claim is.
But equally now, we're starting to see claim forms where you've got a part attachment that is getting to the 10, 12, 18, 20-odd pages, and that also then takes the time of identifying we're moving away from what the key points of agreeance there to what are the actual complaints being raised in the claim, and you still get the same issues that an employee's understandably relying upon what AI has created, but then perhaps not understanding the concepts and what they're actually trying to get across.
And I suppose on the flip side of it, there's an element for some grievances, AI use is probably quite helpful in a way that if an employee is really struggling to articulate what their concerns are, if it's used in the right way with the right questions being asked and the right questions coming back from AI, it can take what could be a bit of a rambling grievance into something a bit more structured and a bit clearer for employers to use.
But I think the example you gave earlier in terms of what you've road tested using Copilot and all the different questions, there is that risk there that you're almost needing the employee to understand if I don't need to answer yes to that question because I'm happy with what I've got now, rather than just keep ticking yes and getting a longer and longer grievance.
So there's an element there of if it's used properly, AI probably can be helpful in drafting a grievance. Yeah. Agreed. And actually, could potentially be helpful in if someone produces a 20-page grievance, could be helpful to the HR team in using AI, can you distil this into three or four key points?
However, we are all relying upon technology in that respect, and it can miss things, and there is a danger there that you... Actually, the key point of the 20-page grievance was buried away at the bottom of page 18, and it misses it. So, I think we're all living in a world where there's a lot of use to- that we can make of AI and in these types of scenarios, but I think we just have to recognise the dangers that it brings.
But as you say, if it is helping an employee who is genuinely aggrieved to articulate what they are aggrieved about rather than what we perhaps saw in the past, which was just reams and reams of complaints, yeah, great, if it can articulate that. But it goes the other way as well, and so we're all grappling with this.
We are as lawyers, HR professionals are, etc., and employees are. They've just discovered it as well. They're not getting it right, and that's causing an awful lot of difficulty and time-wasting in some respects for HR teams. Yeah, I'd agree with that. I think at this point we'll just take a very quick break here, and when we come back, we'll have a quick look at some points that we might start to see in terms of AI use in processes like grievances and how we think employers might start to react.
So welcome back to part two of this Mills & Reeve employment podcast. Richard, we've had a quick chat about substance and where AI might be impacted on that in grievance, but I suppose there's an element as well, isn't there, in terms of process and the complaints we're seeing being brought up about that as well?
Yeah, I think that's certainly something I'm noticing is that I think, again, going back to basics, ultimately all the AI we're talking about, they're large language models, and so they rely upon the information that they have from all sorts of millions of sources. And so what I sometimes see is that there's quite a lot of focus in these grievances on the actual process that's followed.
And yes, we're lawyers, process is important, etc., but you do sometimes miss out on what the substance is. And so to give some examples, I'm increasingly seeing grievances that say, "Actually, you never sent me the minutes of this meeting," and, "Actually, you didn't do this," and, "Actually, in line with your policy, you should have done this."
And those may be valid complaints, they may not be, but that almost dilutes the actual substance of the complaint. And that's where sometimes you almost want to say to the employees, "Look, you are watering down... You might have a really valid point about something that's happened. You might not have got the promotion compared to a female colleague or a male colleague or someone older or younger or whatever.
You might have validity on that. But if you start then saying, 'And you didn't send me the minutes of the notes,' and, 'Actually you didn't let me have a companion,' and, 'You didn't do this,' and all these procedural matters, you are diluting it." So I think there's a risk there that- Probably more of a risk for the employees, but for employers, you can lose focus on what the actual complaint is.
Which I guess brings me back to that initial point, which is sometimes you just have to sit down with people and say, "What is the issue?" Yeah. "And what do you want out of it?" And it- it's one of those matters as well. I think it's years old that when it comes to grievances, quite often a lot of grievance policies will almost have the, "We estimate it could be five days for us to complete this process, but it might take longer."
And I think what I'm seeing a lot with the AI grievances, and that really comes to focus, that if it takes longer than what estimated time it might be, then you sometimes get a follow-up complaint almost, don't you? That AI's gone back in and said, "You've now breached this and this, your policy, ACAS guidance, because you've not done this as quickly as possible."
And then it gets into the further process territory where you're then almost getting a little bit circular, that it's taken us so long to investigate something, but then actually because you're complaining about how long it's taken, we're adding that on. It's just making the process even longer as well.
So time management, I think of a lot of processes where there's AI involved can almost go a little bit skewwhiff. I find- Yeah ... that a lot of that is coming from the procedure-based complaints that AI's generating a bit of. I think there's another point around that as well actually, because I think all of that's valid, but we've focused quite a lot on grievances, but actually sometimes we're seeing it just in responses to internal emails.
Someone, a, a manager has sent you an email saying, "You were late for your shift," or, "I've noticed you didn't do this," or, "You have done this." And AI being used to generate responses to that. And again, we live in quite an angry world at the moment where everyone seems adversarial in everything that they do.
It's actually generating an awful lot of noise around general interactions within the workplace. And I think it is something that people have to be mindful of. And I'm not saying everyone's doing it at all, but we know sometimes the types of employees who will get an email that they're unhappy with and get AI to generate it, and it's not necessarily a grievance, and it still is that- Exactly as you've just described, Michelle, that time management in terms of adding additional time, whereas there wouldn't have even been a reply previously, and suddenly the manager's having to go to HR and say, "I've had this, they seem aggrieved," etc.
So It might save time in terms of creating the actual, the, the original letter. It doesn't save time in terms of the process. Yeah. And I just find it interesting on the flip side where it is something like that, where it's just day-to-day conversation almost, as you say, of things being raised, AI generating a response.
But then if the manager at the same time also uses AI, and you think almost you've got two AI-generated responses," then the AI's arguing between itself almost in terms of, "This is what I think is the issue," and then they get a response from the manager with the same points as well. It feels like there's a, an element of trying to traverse a minefield a little bit, isn't it?
It feels like it's only going to become more prevalent in the future as well, and it perhaps is going to stray from where it might feel a lot confined at the moment to complaints and formal processes. As you say, it might start trickling into day-to-day communications and just generally getting on with the job, which can cause a bit of a headache as well.
Yeah. Absolutely. You're right. We could end up with robots talking to one another, but if everyone's actually getting on with the day job, then maybe that would be helpful. Maybe some people would quite like that. There's that element, isn't there? It certainly feels like something that's here to stay and as you say, to probably grow as well.
I suppose as a final thought, in terms of from an employer perspective and how they might be able to manage or address future use, is there a view in terms of policies or anything they could put in place? And it's almost difficult to manage, isn't it, in a way, because if it's used on personal devices or anything like that, is there anything actually employers can do to almost narrow the scope where AI can be used or how they can manage it?
Yeah, I think you touched on it there that we've all got Copilot, Gemini, ChatGPT equivalents on our personal devices. So I don't think it's one for any kind of technology or AI use internal policies, because often these things won't be done on your policies and people can go and access it on their personal devices anyway.
I think that we are... It certainly is something that, in the grand scheme of things, is very new, and it's going to continue to build, and we'll get more and more focus on it as we see. We're always talking about in employment law certain societal trends where actually the law takes a little while to catch up with it.
I think on policies, it's definitely worth probably reviewing grievance policies just to ensure that you cover off probably more so than anything the actual meeting and the importance of the meeting and the sitting down face to face, whether that's on Teams, whether that's in person. But increasingly now, that's going to become more important than the actual letter.
So I don't think the framework is still there with the ACAS code and with people's policies. It's not radically altering that. But I think it's probably worth reviewing some of those policies just to ensure that points around we will seek to meet with you to distil your complaint down into... You don't need to limit it to a certain number of points, but we will...
To fully understand the nature and the essence and the substance of your concerns, etc., and also what you want out of it. I think that's ultimately there are sometimes questions about whether grievance is the right word. It's intended to be workplace resolution. Grievance sounds more adversarial by its very nature, and actually we have lost sight slightly even before AI of the fact that it's intended to be there to resolve workplace issues.
So a natural question is, "What do you want out of this?" rather than just throwing a load of complaints at us. So I think that really it's going to place more importance on the actual meeting and the actual discussion with the employees rather than what is documented in writing, which from a lawyer saying what's documented in writing is less important sounds slightly counterintuitive, but it's going to put a real focus on the actual meetings and the human interactions.
Yeah, I'd agree with all that, and I feel like that's a great final thought for us to end on for this podcast as well. But thank you to everyone who's joined us, and please follow and subscribe to the podcast if you haven't already done so. If you'd like to keep the conversation going and share your own thoughts, please feel free to reach out to Richard and I on LinkedIn or through our contact details on the Mills & Reeve website.