Generative AI is rapidly becoming part of everyday working life. Employees are using it to draft emails, summarise information and prepare documents, while managers are increasingly turning to it for help with everything from meeting notes to difficult conversations.
Used well, it can save time. But Employee Relations (ER) is one area where the risks of getting it wrong are particularly high.
Grievances, disciplinaries, investigations, absence issues and workplace disputes are rarely straightforward. They involve context, judgement, emotion and often highly confidential information. When either an employee or a manager relies too heavily on AI, what initially looks like a helpful shortcut can make an already difficult situation more complicated.
This is becoming a live issue for employers. Acas is currently consulting on updates to its Code of Practice on disciplinary and grievance procedures and has specifically asked for views on the benefits and risks arising from the use of AI in these processes.
When employees use AI in ER cases
It is increasingly easy for an employee who is unhappy at work to ask an AI tool to:
- draft a formal grievance
- strengthen their allegations
- identify potential breaches of employment law
- prepare questions for a disciplinary hearing
- write an appeal
- respond to an investigation
- turn a relatively straightforward complaint into a lengthy formal document.
There is nothing inherently wrong with an employee getting help to organise their thoughts or communicate more clearly. The difficulty comes when the AI-generated content starts to overtake the employee’s actual experience.
A relatively simple workplace disagreement can suddenly arrive with HR as a multi-page grievance filled with legal terminology, numerous allegations and references to duties or legislation that may not actually be relevant.
The CIPD has highlighted this emerging problem, noting that generative AI can contribute to longer and more complex grievances and correspondence, sometimes including irrelevant or inaccurate points. The result can be considerably more work for HR and managers to establish what the employee is actually alleging and what needs to be investigated.
It can also escalate the tone.
An employee might ask AI to make their complaint “stronger” or “more formal”. The resulting document may describe an uncomfortable conversation as bullying, discrimination, harassment or victimisation without the employee fully understanding the meaning or implications of those terms.
That does not mean the concerns should be dismissed. Far from it. Employers still need to establish the facts and take concerns seriously. But it can make it harder to identify the core issue and resolve it quickly.
There is another problem: accuracy.
Generative AI can confidently produce information that is wrong. An employee may therefore submit arguments based on legislation, company policies, case law or supposed employment rights that do not apply to their situation.
What started as a genuine concern can become buried beneath AI-generated argument.
The confidentiality risk
ER cases frequently involve sensitive information about several people.
Employees using publicly available AI tools may copy and paste emails, witness statements, meeting notes, medical information, names or allegations about colleagues into them without considering where that information is going.
Managers can make exactly the same mistake.
This creates an important data protection and information security issue. The ICO makes clear that data protection requirements apply to personal information used within AI systems, including information entered by users through prompts.
Organisations therefore need clear rules about what employees and managers can put into external AI platforms.
Managers face an even greater responsibility
The potential consequences become more significant when AI is being used by the person responsible for managing the ER process.
Imagine a manager asking an AI tool:
“I’ve interviewed these three witnesses. Who do you think is telling the truth?”
Or:
“Based on these investigation notes, should I dismiss this employee?”
Or even:
“Write me an outcome letter confirming that this employee committed gross misconduct.”
At that point AI is no longer simply helping with administration. It is influencing judgement.
That is dangerous territory.
A fair ER process requires the employer to consider the evidence, understand the circumstances, listen to the individuals involved and reach a reasonable conclusion. Acas guidance makes clear that employers should carry out a reasonable investigation before making decisions in disciplinary and grievance cases.
A line manager cannot outsource that responsibility to an algorithm.
The danger of AI-generated outcome letters
One of the most tempting uses of AI is drafting correspondence.
There may be a role for it here. A manager could potentially use an approved company AI tool to help improve the structure or clarity of a letter.
But there is an important distinction between:
“Help me express the decision I have reached clearly.”
and
“Tell me what decision I should reach.”
An AI-generated disciplinary or grievance outcome might sound polished and authoritative while containing assumptions the manager has never actually considered.
It might introduce findings that were not established during the investigation. It might exaggerate the evidence. It could use inconsistent reasoning or reference policies incorrectly.
A manager should never send an ER outcome they could not personally explain and defend.
If challenged, “the AI wrote it” is not a satisfactory explanation for how an employment decision was reached.
Bias does not disappear because a computer is involved
There can also be a tendency to assume AI provides an objective answer.
It does not.
AI systems work from data, prompts and patterns. They do not understand an organisation’s culture, the personalities involved, previous conversations or the nuances of a working relationship in the same way an experienced manager or HR professional can.
Nor should employers assume that introducing an AI tool removes the risk of bias.
ICO guidance around automated workplace decisions emphasises the importance of meaningful human oversight and giving workers an opportunity to challenge decisions that affect them.
In sensitive ER matters, human judgement is not an inconvenience to be automated away. It is an essential part of making a fair decision.
So should AI be banned from Employee Relations?
Probably not.
There are legitimate ways AI can support an ER process.
It might help someone structure their thoughts ahead of a difficult conversation. It could help a manager prepare a neutral meeting agenda, identify questions they may want to explore or turn their own notes into a clearer format.
Within an organisation’s approved systems and appropriate data protection controls, it could also reduce some of the administrative burden around ER.
The important word is support.
AI should support the person managing the case, not replace their judgement.
Employers need some ground rules
As the technology becomes commonplace, simply telling employees not to use AI is unlikely to be enough.
Organisations should instead consider establishing clear expectations around its use, including:
- what information must never be entered into public AI tools
- whether AI can be used to prepare ER correspondence
- whether employees should disclose significant use of AI when submitting formal documents
- how managers should check AI-generated content for accuracy
- circumstances where AI must not be used to make or recommend employment decisions
- the requirement for meaningful human review of anything produced using AI
- appropriate use of secure, organisation-approved AI platforms.
Managers also need training. Someone who understands how to conduct a fair investigation will be much better placed to recognise when an AI-generated suggestion is inappropriate than someone who is relying on the technology because they lack confidence in the process.
Keep the human in Human Resources
AI can process information extremely quickly. What it cannot do is take responsibility for an employment decision.
Employee Relations often requires managers to deal with ambiguity, conflicting accounts, damaged relationships and difficult emotions. There may not be an obvious answer.
Good ER practice involves asking the right questions, listening carefully, testing evidence and exercising judgement.
Those are human responsibilities.
As AI becomes embedded in everyday work, employers will need to find the right balance: taking advantage of technology without allowing speed and convenience to undermine fairness, confidentiality or common sense.
For organisations dealing with complex Employee Relations cases, independent HR support can provide an additional layer of experience and objectivity. Hunter Adams supports businesses across the UK with investigations, disciplinaries, grievances and wider ER matters, either alongside existing HR teams or as an outsourced HR partner.
If an ER case is becoming difficult to navigate, speak to the Hunter Adams team about how we can help.
Find our more about our Employee Relations services here >> https://hunteradams.co.uk/hr-consulting/employee-relations-services/
