“AI-related grievance” sounds like one thing. It isn’t. Inside that single phrase sit two genuinely different situations, and HR teams that treat them the same risk mishandling at least one of them.
In the first, an algorithm helped produce the outcome an employee is complaining about, a shift pattern, a performance flag, a rejected application. In the second, the employee has used AI to help write or build the grievance itself, regardless of what it’s actually about. Both now show up regularly in UK workplaces. Neither fits neatly into a grievance process built around a single human-authored complaint about a single human decision.
This piece breaks down what each type actually looks like, how each typically moves through a grievance process, where standard HR procedures fall short for both, and what to do differently.
Explore: AI Grievances and Employment Tribunals: A Guide for HR Teams, for the fuller picture of how AI-related disputes reach a tribunal.
The Two Kinds of AI-Related Grievance
The first kind starts with a decision. Recruitment scoring, shift scheduling, performance monitoring, an algorithm was somewhere in the chain that produced an outcome the employee is unhappy with. The employee’s complaint is about that outcome. AI’s involvement may not even be visible to them at the point they raise it.
The second kind starts with a tool, not a decision. The employee reaches for ChatGPT or a similar tool to help articulate a grievance, appeal a disciplinary outcome, or challenge a redundancy, regardless of the underlying issue. What lands on HR’s desk looks different: longer, more formally worded, often citing law or case references the employee wouldn’t ordinarily reach for unaided.
Both are now common enough to be routine rather than exceptional:
- 83% of HR directors had worked with a business facing issues related to employee AI use in the past twelve months
- 95% had specifically encountered AI used to raise a grievance or progress a dispute
(The HR Dept’s survey, reported in People Management.) This isn’t an edge case HR might encounter once. It’s close to universal.
Treating these two situations as the same problem misses what each one actually requires. One calls for technical and procedural questions about how a system reached its output. The other calls for reading past the packaging to find the actual underlying complaint. Conflating them means answering the wrong question for at least one of them.
Anatomy of a Decision-Origin Grievance
This type rarely starts with the words “an algorithm treated me unfairly.” It starts with something concrete: a shift the employee didn’t expect, a performance rating that seems wrong, an application that went nowhere without explanation. The complaint is about the outcome. What produced it is often invisible to the employee at the point they first raise a concern.
That invisibility shapes how these grievances typically escalate. An informal query usually goes to a line manager who has no more insight into the system than the employee does. If it becomes a formal grievance, it’s still usually framed around the visible outcome, not the process behind it. The explicit question, was AI involved in this, and how, tends to surface only at escalation, sometimes for the first time in the entire process.
Manjang v Uber Eats UK Ltd is the clearest real example of this pattern. A driver was suspended after repeatedly failing a facial recognition check. His complaint wasn’t framed as “the algorithm is biased,” it was framed as an unexplained, unreviewed suspension. The discrimination and procedural fairness questions only became explicit once the case reached a tribunal.
The structural issue this exposes is straightforward: standard grievance procedures assume a human decision-maker who can be asked to explain their reasoning directly. That assumption doesn’t hold when a system produced or shaped the decision, and nobody has been assigned to answer for it.
Anatomy of an AI-Assisted Grievance
This type is usually easy to spot once you know the pattern: unusually long, formally worded, often citing statutes or case law the employee wouldn’t typically reach for on their own, and frequently disproportionate in tone to the substance of the complaint.
It’s also now common enough to be a genuine HR trend rather than an occasional oddity. Irwin Mitchell’s survey of 200 HR professionals found:
- 60% suspected AI involvement in at least one grievance they’d handled
- 1 in 3 HR managers reported involvement in a tribunal claim where they believed AI had been used to prepare the employee’s case
This has already moved well past the drafting stage. The HR Dept’s survey found 78% of AI-generated grievances relied on inaccurate information or misrepresentations, with 11% containing substantial factual errors. The most striking illustration comes from Thrings: a grievance letter that cited nine legal cases, of which only two were real.
None of this changes what HR is required to do. The ACAS Code of Practice still requires a reasonable investigation of every grievance, regardless of how it was written. What changes is how carefully that investigation needs to read past the formatting to find the actual complaint underneath it.
Where HR Processes Break Down for Both Categories
- Decision-origin grievances: the gap shows up in documentation, not just ownership. Even where someone could technically explain a system’s output, there’s rarely a record of the human review that happened at the time the original decision was made, so there’s nothing to point back to once a grievance arrives. ICO’s Recruitment Rewired findings reflect a version of the same gap from the regulatory side: many employers believe they have meaningful human involvement in their recruitment decisions when, in practice, decisions were made by the system alone, and where human involvement does exist, it isn’t always applied consistently to every candidate.
- AI-assisted grievances: the gap is about readiness. 52% of respondents said AI-drafted grievances were harder to resolve (Irwin Mitchell), citing overly formal language and a lack of personal context. Most HR managers surveyed had received no training on how to approach these differently.
The two gaps share a root cause. Grievance procedures were built around a single human-authored complaint about a single human decision. Neither category fits that assumption cleanly, one because the decision-maker wasn’t fully human, the other because the complaint wasn’t fully human-authored either.
What This Means for How HR Should Respond
- For decision-origin grievances: assign clear ownership for explaining how AI-influenced decisions are made before a complaint arrives. Document the human review point at the time the original decision happens, not reconstructed afterward, by which point it’s too late to show one existed.
- For AI-assisted grievances: strip away the formatting and legal-sounding language, identify the specific incident the employee is actually pointing to, and investigate that under the same ACAS Code process used for any other grievance. The drafting method doesn’t change what’s owed to the employee, it changes how much care is needed to see past the packaging.
How Avado Can Help
Whether a grievance originates from an algorithmic decision or arrives shaped by AI-assisted drafting, the underlying skill HR needs is the same: knowing how to investigate fairly, ask the right questions, and reach a defensible outcome. Avado’s HR Compliance for Managers course, presented by employment law specialist Amanda Chadwick, builds exactly that judgement, covering disciplinary and grievance essentials alongside the wider legal landscape managers now operate in.
Explore HR Compliance for Managers and make sure every grievance gets a fair, defensible response, however it arrived!
