Employee grievances have become a whole different ball game since AI became an everyday tool for the average human. 15 pages, 10 different allegations, several references to legislation you've never heard of and a demand for a response by 4pm Tuesday. Welcome to employee relations in the age of AI.

We've talked before about employees increasingly having what essentially feels like an HR Adviser sitting in their pocket, but a recent Australian workplace case has highlighted a much more interesting problem with AI generated grievances.

What happens when someone actually has a genuine complaint, but AI makes it so difficult to find that everyone stops listening? That's pretty much what happened here, and the fact the complaint was real is exactly what makes this case so interesting.

Deidre De Bruyn brought a workplace bullying complaint against the Body Corporate for the Sanctuary in Australia. This wasn't simply somebody asking ChatGPT whether their manager being a bit grumpy with them on Tuesday constituted unlawful bullying; there were genuine concerns underneath it. The Fair Work Commission found conduct towards De Bruyn that was intrusive, unreasonable and targeted, including monitoring her activities, circulating derogatory comments and publicly criticising aspects of her work.

However, what also came to light was that De Bruyn's correspondence became lengthy, wide ranging, repetitive and accusatory, with the Commissioner stating that the correspondence was “unhelpful” as well as arguing that it was “unsurprising” the employer had, as a result, simply stopped responding to the worker’s lengthy AI generated correspondence. 

The problem wasn't that AI had created a complaint where there wasn't one; it was that AI had helped create so much noise around a genuine complaint that the actual issue became increasingly difficult to hear. The volume and tone of the correspondence meant that the very people De Bruyn needed to engage with effectively started to switch off. You therefore end up with quite an extraordinary situation where somebody has genuinely experienced concerning behaviour, but their attempts to communicate about it become so overwhelming that the substance of their complaint gets buried underneath the way it's being presented. That should make leaders and HR professionals pay attention far more than another story about ChatGPT getting the law wrong.

In this case it got to the point where the Commissioner put some very clear boundaries around what happened next, ordering that future communication between the parties had to be brief, accurate and respectful, and that AI tools were not to be used to prepare their correspondence at all. That's a pretty significant step, and it matters because the issue wasn't simply that AI had made the process annoying or overcomplicated - it had become so dominant in the communication that the Commissioner effectively had to remove it from the equation altogether. You can read the full case here.

For years, there's been a fairly obvious imbalance in workplace disputes. The employer usually has HR support, maybe even an employment lawyer, and they're experienced with handling grievances. The employee may have had Google and a friend who once had a challenging situation with an employer and therefore offers their pennies' worth. AI has now completely changed that - an employee can now describe what happened at work and, within seconds, receive something resembling legal advice, a breakdown of possible claims and a lovely formatted grievance letter.

That isn't automatically a bad thing - there are some genuinely positive implications in employees being able to use AI in the right way. Someone who struggles with writing can organise their thoughts. Someone who speaks English as an additional language can communicate more confidently. Someone who is neurodivergent may find it easier to structure what happened. Someone intimidated by HR terminology can ask what it actually means. Someone who knows something feels wrong but doesn't know how to articulate it can use AI to help make sense of the situation. So if used well, AI can actually make workplace processes more accessible, which of course matters.

At Fresh Seed we're already seeing a trend in hefty AI assisted grievances being submitted, with each subsequent response quite obviously being put back through ChatGPT or Claude AI. This recent case isn't really about whether employees should be allowed to use AI though; it's about the point at which AI stops helping somebody communicate and starts getting in the way of what they're actually trying to say.

AI is extremely good at sounding certain and it's also extremely good at telling you what you want to hear. Tell it about something that happened at work and ask whether your employer has broken the law, and it can produce an impressively alarming list of possibilities. Ask it to make your email more assertive and suddenly you're writing like you've instructed a barrister; ask what documents you're entitled to and five minutes later you're requesting records nobody knew existed. The language becomes more sophisticated and the legal terminology more impressive, but (and this is crucial) the person's actual understanding of their situation hasn't necessarily changed at all, which is where it becomes dangerous in employee relations.

A fairly normal disagreement can suddenly become:

“ I’m formally requesting immediate confirmation of the organisation's compliance with its statutory obligations and reserve all rights in relation to potential claims…” 

When what the employee actually means is:

“I'm really upset about what happened and I don't feel anyone has listened to me.”

Those require very different conversations, and there's an irony here too: making a grievance sound more serious can sometimes make it harder for the serious bits to be heard.

If every email is eight pages long, every disagreement becomes a potential statutory breach and every response generates another AI assisted rebuttal, eventually the humans at either end stop talking to each other. The employer sees an enormous legal document rather than an upset employee and the employee sees a defensive corporate response rather than another human being trying to understand what happened. Somewhere underneath paragraph 47 is the thing everyone should actually have been talking about in the first place.

There's another side to this that employers need to be really conscious of. You receive a 10 page grievance littered with legal terminology, then another email arrives, then another allegation, then 14 questions about your investigation process followed by a demand for copies of seven policies. Naturally it's overwhelming, and a very understandable reaction might be: “This is obviously extremely serious and we need lawyers involved.”

The lawyers respond, the employee puts that response into AI, AI produces the next response, the lawyers respond to that and the employee puts that back into AI. Suddenly the original disagreement about how somebody was treated during a team meeting last Tuesday has turned into a massive litigation exercise costing £££££ and consuming the time and energy of half the senior management team, while nobody has actually had a useful conversation.

There's another very human reaction too, which is that people start switching off such as in the case above. By page eleven the person reading it is skimming, by the fifth email they're frustrated and by allegation number 17 they're starting to assume the whole thing is ridiculous and start pushing it to one side. 

That's the dangerous bit, because irritating correspondence doesn't mean there isn't a legitimate complaint underneath it. A badly articulated grievance can contain a serious allegation, an employee who has misunderstood the law can still have been treated badly, and an AI inflated grievance can still contain something that absolutely needs investigating. De Bruyn's case demonstrates exactly why employers need to remember that.

If you've received a million pages of allegations from an employee, don't automatically assume the employee understands or even agrees with every word. It's best practice to ask them about it during the grievance meeting and bring the conversation back to something simple like:

• There's quite a lot here. Can you talk me through what happened from your perspective?
• Which of these concerns is most important to you?
• What impact has this had on you?
• What are you hoping will happen as a result of raising this?

Questions like these get you talking to the human again and help you dissect the AI bumf to get to the real issue of what's actually going on.

Something we think is going to become increasingly important as an ER skill for HR professionals and managers is not confusing polished language with the strength of the case. Something that's beautifully written and cites six pieces of legislation isn't necessarily a good grievance, just as something that's repetitive, frustrating, AI generated and 14 pages long isn't necessarily a bad one.

If an employee submits twelve allegations, the employer still needs to work out what actually happened, what evidence exists and what genuinely requires investigation; the skill isn't spotting whether ChatGPT wrote it, but finding the human issue underneath it.

The next part's also important - please don't respond with AI generated HR advice. There's something slightly ridiculous about employers becoming outraged that employees are using AI to write grievances while HR teams are simultaneously putting those grievances into ChatGPT and asking it to write the response.

AI absolutely does have a place in HR. It might help organise a timeline of events or identify themes across a large amount of information, assuming you've got your data protection and confidentiality arrangements sorted, and it might also help you sense check whether a letter sounds unnecessarily defensive or turn your rambling notes into something somebody else can actually follow. But deciding whether somebody has been bullied or discriminated against? Absolutely not, because these are human situations requiring context, curiosity, judgement and occasionally the ability to recognise that two people have simply really annoyed each other.

So, should employers ban AI from grievances? We don't think so. There are perfectly legitimate reasons why somebody might use AI to help communicate a workplace concern, and a blanket ban could disproportionately affect people who use these tools for accessibility, language or communication support, as well as being another challenge entirely to police. Instead, employers should think about setting sensible boundaries around how AI is used rather than pretending it isn't being used at all.

For example:

• Don't input confidential company, colleague or client information into unapproved public AI tools.
• AI can help structure your thoughts, but the account you give should genuinely be yours.
• Don't include allegations simply because an AI tool suggested they might apply.
• Check legal information before relying on it.
• Be prepared to explain the concerns you've raised in your own words.
• Remember that you remain responsible for anything you submit or send.

That's considerably more useful than NO CHATGPT IN GRIEVANCES.

Employees aren't going to stop using AI and neither are employers, so the organisations that handle this well probably won't be the ones with the strictest AI policies, but those whose managers and HR teams know when technology is helping people communicate and when it's getting in the way of the actual conversation. Perhaps the biggest lesson from cases like De Bruyn's isn't that AI can manufacture an enormous grievance; it's something slightly more uncomfortable, which is that AI can also make a genuine grievance easier to ignore, and that's a much bigger problem.

So, one question to leave you with: if an employee submitted an AI generated grievance to your business tomorrow which was 10 pages long, packed with legal terminology and contained allegations your manager had never even heard of, would they know what to do next?

If the answer is “erm…” then that's probably a conversation worth having with us before ChatGPT joins your next grievance meeting.

At Fresh Seed, we help businesses build people practices that work in the real world, including the increasingly weird world where AI is sitting on both sides of the employee relations table. If you want to sense check whether your AI guidance, grievance processes and manager training are ready for that world, reach out to us using the button below.