When AI escalates: grievances, conflict & keeping humans in the conversation

How AI changes the grievance process

People have always sought help drafting difficult letters — from friends, union reps, solicitors. What has changed is the speed, the availability, and the nature of the help. AI tools are designed to be useful and agreeable. Ask one to help you write a grievance and it will strengthen your framing, sharpen your language, and — because it is trying to be helpful — validate the position you have already taken. It will not push back. It will not ask whether you have considered the other person's perspective. It will produce a document that reads with confidence and authority, even if that confidence is not quite warranted.

The same thing happens on the HR side. A response drafted with AI assistance may be precise, measured, and legally cautious — but it can also feel cold and evasive to the person who receives it. The humanity drains out of the exchange on both sides, and what started as a workplace difficulty begins to look like the opening moves of a dispute.

The sycophancy problem

One of the most important things to understand about AI tools — whether or not you can name the technical reason — is that they are built to be agreeable. They validate your framing, agree with your interpretation, and reflect your view back at you in stronger language. This is sometimes called sycophancy: the model tells you what you want to hear.

In a grievance context, this matters enormously. If you go to an AI feeling wronged, it will help you feel more wronged, and more certain about it. If HR goes to an AI feeling that a complaint is overblown, the response may come back subtly dismissive. Neither output reflects a balanced read of the situation — it reflects the user's starting position, polished and amplified. Critical thinking is not optional here. It is the thing that makes the difference between AI as a useful drafting aid and AI as an escalation engine.

The hallucination risk in formal complaints

AI tools can produce text that looks authoritative but is not accurate. In a grievance context, this might mean:

• Legal citations that do not exist, or cases that have been mischaracterised.
• References to policies, legislation or regulations that are wrong for your jurisdiction or out of date.
• Confident assertions about employment law that a qualified adviser would not make.
• Framing that overstates the severity of what happened, because the AI is optimising for a persuasive document.

Neither party should accept AI-generated legal framing at face value. Verify any specific claim. If a letter cites a statute or a tribunal decision, look it up before acting on it. The fact that it appears in a formally structured document does not make it correct.

What gets lost when AI does the talking

Most workplace conflicts, when they surface as formal grievances, are not primarily legal disputes. They are human situations: someone felt overlooked, disrespected, excluded or mismanaged. The formal process exists to resolve those situations — but it can also entrench them. When AI steps in on both sides, the human signal gets buried under legal language. The question "what actually happened, and how do we move forward?" gets replaced by "what can be proven, and who bears liability?"

What gets lost is nuance, context, tone, and the possibility of repair. Reconciliation requires both sides to feel heard as people. That is very hard to achieve through documents that neither side actually wrote.

Keeping humans in the conversation

Whether you are an employee, a manager, or an HR professional, there are practical things you can do to keep the human element central:

• Treat AI output as a draft, not a finished document. Read every sentence and ask: is this what I actually want to say? Does it reflect what genuinely happened? Would I say this to the person's face?
• Verify before you rely. Any legal citation, policy reference or regulatory claim should be checked independently. Do not send a document that makes assertions you cannot personally stand behind.
• Resist the pull towards position-taking. Formal language locks people into stances. If the situation can be addressed informally — a facilitated conversation, a mediation, a direct discussion — that is nearly always preferable to exchanging documents.
• Name what is happening. If you suspect an AI-drafted letter has inflated the framing, it is reasonable to say so — calmly, without dismissing the underlying concern. The complaint may be entirely sincere even if the language is not quite the person's own.
• Pick up the phone. Text on a screen cannot convey tone, regret, willingness, or good faith the way a voice can. Before the process hardens, try a conversation.

For leaders and HR: what this means in practice

This is not an argument against employees using AI, or against HR using it. It is an argument for awareness. People will use the tools available to them — that is human nature, and AI tools are genuinely useful for people who find formal writing daunting or who struggle to articulate something in the heat of the moment.

What leaders and HR teams can do is build a culture and a process that makes de-escalation easier than escalation. That means early informal resolution routes, trained managers who can hold difficult conversations well, and a commitment to understanding what someone is really trying to say — not just what their letter appears to claim.

It also means being honest with yourselves about how your own use of AI tools might be shaping your responses. If your reply to a grievance was drafted with AI assistance, read it with the same critical eye you would want the other side to apply. Does it actually address the person's concern, or does it just close down the argument?

AI, belonging, and the future of work

The AI grievance loop is one instance of a broader challenge: AI tools are increasingly shaping how we communicate at work, and not always in ways that bring people together. Explore more on this theme in the guide to AI, belonging and the future of work, or hear how I explore these ideas through the speaking topic AI, Belonging & The Future of Work. For a wider view, browse more guides or listen to the Inclusion Bites podcast.