AI Can Help With a Redundancy Process, But It Shouldn't Do the Thinking for You
Someone in HR told me something recently that I keep coming back to.
They'd been dealing with redundancy situations where managers had used AI to help write the business case. What came back was long, detailed and very polished. The trouble was, it didn't really understand the business or the team.
I can see why it's happening.
Writing a redundancy business case is hard. The manager might be under financial pressure, pulling together information from different parts of the business, and worrying about how it'll affect people they know. Faced with a blank page, asking an AI tool for help seems sensible.
But a long document can make the thinking look a lot more complete than it really is.
The business case needs to come from the business
AI can come up with a familiar-sounding reason for almost any restructure: economic pressure, duplication, efficiency, changing customer demand or simplifying the organisation.
Any of those might be true. On their own, they're not a business case.
The real work is much more specific. What's actually changed? What evidence backs that up? Why isn't the current structure working any more? What other options have you looked at? How does the new structure fix the actual problem? And what assumptions still need testing?
You can only answer those if you know the organisation. And someone has to take responsibility for the answers.
If a manager types in a short prompt and gets several pages back, the risk isn't just that some of it's wrong. It can add reasons the manager never thought of, hide gaps in the evidence, or make a proposal sound like a done deal before consultation has even started.
Consultation should be a genuine chance to look at ways of avoiding redundancies, reducing the numbers and softening the impact on people. It's not just a meeting to explain a decision that's already been made, and the government's own guidance on redundancy consultation is clear on that.
Why the risk is going up
The CIPD's Summer 2026 Labour Market Outlook found that 17% of employers expected staffing levels to fall over the following three months.
At the same time, getting it wrong is becoming more costly. Since April 2026, the maximum protective award for failing to consult properly on collective redundancies has doubled from 90 to 180 days' pay. From January 2027, people will only need six months' service to claim ordinary unfair dismissal, down from two years, and the cap on compensation is being removed. New organisation-wide collective redundancy rules are also due in 2027, although the detail is still being worked out. I've covered the wider changes in my guide to the Employment Rights Act 2025 for small businesses.
That should push businesses towards earlier planning and better decisions. Not just longer documents.
So where can AI actually help?
I do think AI can be useful here, as long as the thinking and the accountability stay with people.
It could help a manager put checked information into a timeline. It could compare different versions of a proposal and spot where they don't match. It could read a draft and point out jargon, repetition and the bits that are hard to follow.
It can also be a really good challenger. You could ask it things like:
Which points in this draft aren't backed up by evidence?
What reasonable alternatives haven't been covered?
What questions might an affected employee or rep ask?
Does the new structure actually solve the problem described?
Which parts sound like the decision's already been made?
To me, that's far more useful than asking it to "write a business case for removing three roles".
In the first case, AI is helping people test and explain their own thinking. In the second, it's being asked to do the thinking for them.
It can help with the work around the decision too
There are some admin jobs where using AI carefully could free up time for the human side.
It could help create scenarios for manager training, draft an FAQ from information you've already approved, or group anonymised consultation questions into themes. It could check whether a letter covers the key dates, where to get support and who to contact.
It still needs proper oversight, though. Information about employees, their health, performance or protected characteristics shouldn't go into an AI tool that hasn't been approved for it. Everything it produces needs checking against the facts, your process and current legal advice. And AI should never choose who's selected for redundancy or make an employment decision.
The ICO's guidance on automated decisions affecting workers makes a good wider point: human involvement has to be meaningful. The person checking the output needs the knowledge and authority to challenge it, not just accept whatever the system suggests.
The human side can't be generated
Nobody likes making redundancies. Businesses can be under very real financial pressure and decide that difficult changes have to happen.
But people will remember how they were treated. Whether the reasons made sense, whether their questions were listened to, and whether anyone recognised what it meant for them personally.
That means thinking beyond the paperwork. Acas recommends thinking about support for employees at risk, managers, the people leading consultation, employee reps and the people who stay. Counselling, outplacement, help finding work and a clear plan for the future can all play a part.
AI might help you prepare some of that material. But it can't sit with someone in a hard meeting, really listen, or show compassion and judgement.
Questions I'd ask before using AI
Before using AI anywhere in a redundancy or restructure, I'd want clear answers to these:
What exactly are we asking it to help with?
Is AI right for that job, given the impact on people?
What information will go into it, and is the tool approved for that?
Who's checking every fact and assumption?
Could what it produces make the proposal look like it's already decided?
Who has the knowledge and authority to challenge or reject what it says?
How will we record the real human reasoning and decision?
Would we be happy explaining how we used AI to an employee who's affected?
That last one is the one I find most useful.
AI can make a careful process easier to manage. It shouldn't make a weak process look more convincing. And in a decision that affects people's lives this much, it should never stand in for evidence, accountability or kindness.
If you're planning a restructure and would like help with the process, or you want to work out how AI should (and shouldn't) be used in your HR team, I'd be happy to have a chat. You can get in touch at enquiries@hrnetics.co.uk.
Frequently asked questions
Can managers use AI to write a redundancy business case?
They can use it to help structure, check and simplify a draft, but the reasoning and evidence need to come from the manager and the business. Check every fact, and don't put employees' personal information into an AI tool that hasn't been approved for it.
Can AI be used to select employees for redundancy?
No. Selection should be made by people, using fair and objective criteria. Human involvement in decisions about workers has to be meaningful, not just signing off what a system recommends.
What's changing for redundancy in 2027?
From January 2027, the qualifying period for ordinary unfair dismissal drops from two years to six months and the cap on compensation is removed. New organisation-wide collective redundancy rules are also due during 2027, with the detail still to be confirmed.
Is it safe to put employee information into ChatGPT?
Not into free or unapproved tools. Only use AI tools your organisation has approved for personal information, and anonymise details wherever you can.
Comments