In An AI Centered Workplace Traditional Investigation Training Is Not Enough
70% of surveyed organisations noted an increase in workplace grievances drafted using artificial intelligence tools. 60% of UK HR professionals have recently dealt with workplace grievances they suspected were generated using AI, and 52% said those grievances were more difficult to investigate and resolve.
A mere 12% of employers are confident their managers can handle complex AI-assisted grievance cases, according to a survey of more than 900 HR and business professionals by employment law and HR consultancy WorkNest
These statisticd might sound surprising, but I actually think it highlights a much bigger issue. These issues are compounded by the complainants frequently struggling with mental health or wellbeing issues which caused them to raise the issue in the first place.
Not many employers are taking into account that we’re still asking people to investigate workplace complaints using skills that were designed for a very different world. These new challenges shouldn’t be ignored any longer as AI-assisted employee complaints require a new capability from HR investigators.
They must learn to distinguish between the quality of the writing and the credibility of the evidence.
Rise in AI assisted grievance complaints
Employees are increasingly using AI to help them write grievances – and why wouldn’t they? I’ve even worked with a few clients who have done exactly that and the results are solid.
- AI can produce a polished document, clearly presented and access legal precedents and any local regulations in literally minutes.
- AI can help if the compailant has to file a greivance in a second lanauge or ahve learning differences.
- Others use it to structure their thoughts or make sure they haven’t missed important information.
- Some use it because they’re emotionally exhausted and need help expressing what has happened.
The result is often a professionally written document that can look more like a legal submission than a workplace complaint.
On the downside AI can
- exaggerate legal language
- introduce inaccurate legal references and even hallucinate and make up legal precedents.
- encourage “kitchen sink” complaints containing every conceivable allegation
- make informal resolution more difficult because positions become entrenched before an investigation even begins.
Putting this into perspective to manage the risk involved, in Belgium where I am based, the average employee can create a court ready briefing in hours. Although AI cannot legally represent a client or sign documents as a professional lawyer, it can serve as a highly efficient “co-pilot” for research and documentation ande definitely help handle an unfair dismissal case and minimise any out-of-pocket legal expenses.
The UK is preparing for an increase in AI assisted grievances and expecting an uptick in legal fees to process them as simple complaints escalate using AI. In Australia The Fair Work Commission in a recent case prohibited the use of AI in the grievance procedure noting a spike it its workload saying that “the material generated by GenAI tools can be inaccurate, incomplete, out of date, or just made up”.
When employees start to understand this we will surely see rise in grievances pursued.
More pressure for the investigator
An investigation is there to establish the facts, not judge the quality of the writing. This means that the interview part of the investigation process has to be more attentive, and the first responder, frequently the line manager or HR as well as the official investigator have to be both thorough and empathetic. However there is a small overlooked point, just because something is well written doesn’t mean that its factually accurate, just as a poorly written complaint mean that it’s not valid.
How AI has changed the grievance landscape
Today’s, and tomorrow’s, workplace investigations need expertise that goes well beyond employment law and report writing.
Investigators need a working level of AI literacy: understanding what generative AI can and cannot do, recognising that it may introduce inaccuracies or overstate issues, and learning to separate polished language from verified events. “Validation bias” is deeply embedded in most AI systems and tends to agree with, flatter or reinforce the user’s stated beliefs and assumptions, rather than providing objective, cogent advice if prompted to do so. This results in arguments being advanced by employees which are not always accurate.
Many investigators and first responders will also use AI themselves, organising documents, building timelines, summarising interview notes, and flagging inconsistencies for further review.
This new development in an AI enabled workplace means that first resrponders and investigators will also need sharper interviewing skills to look past the polished veneer of an AI-assisted account. They will need to dig beyond the detailed, sophisticated narratives that still require testing while taking care not to retraumatise the complainant. Investigators must distinguish between facts, assumptions, opinions, and legal arguments, with the investigation driven by evidence, not by the sophistication of the document in front of them.
In short: the skill required to run a good investigatory interview has gone up significantly.
Check out our new Resource Hub: Women and AI
Trauma Informed Investigations
The biggest skill gap is trauma-informed practices. Many grievances involve bullying, discrimination, harassment, or experiences that have caused real psychological distress. Asking someone to relive those experiences repeatedly, through a poorly managed process, risks retraumatising them.
A trauma-informed investigator understands how trauma affects memory, communication, and behaviour. They know how to run a process that stays fair while reducing unnecessary distress, not by lowering the standard of evidence, but by creating the conditions for people to give their best evidence.
Investigators also need a stronger grasp of psychological and psychosocial safety. Increasingly, complaints aren’t about one isolated incident, they’re about patterns of behaviour, leadership practices, and workplace culture. Understanding psychosocial hazards helps investigators see the wider context instead of fixating on individual events in isolation.
And there’s also the question of trust. AI can be a useful tool during an investigation, but it should never decide whether someone is telling the truth, whether a witness is credible, or whether allegations are upheld. Those calls require human judgment, empathy, and context, and AI has none of the three.
New workplace capabilities: toward DE&I 2.0
As AI reshapes how grievances arrive, the capabilities we expect from first responders and investigators need to expand with it. The most effective investigators of the future won’t just know employment law, they’ll combine AI literacy, trauma-informed interviewing, an understanding of psychosocial safety, and sound, neutral judgment.
This is really what I mean by DE&I 2.0: the same commitment to fairness and inclusion that DE&I was built on, extended to cover how people are heard in an AI-enabled workplace, not only if they’re represented in it. It’s the difference between a policy that says everyone has a voice, and a process built to make sure that voice is given a fair hearing once it’s raised, regardless of how it arrived on someone’s desk.
Adopting new technology isn’t the goal in itself, but making sure that as workplaces become more AI-enabled, they also become more human. The real measure of a best-practice investigation isn’t whether AI helped write the grievance, it’s whether the person raising it felt heard, treated fairly, and is able to participate in a process that sought to establish the truth without causing further harm.
