AI-assisted grievances: What’s your biggest challenge?

Holly 

You might wish to do a search on here for this.   An extremely similar post was started some time ago.  I can't remember what the common advice was though.

Thank you David, I will get searching

I haven't come across AI grievances yet (thanking my lucky stars), but for meeting the challenge, I would 100% lean on having an early human conversation before progressing the grievance

thank you for the feedback Sopie, this is helpful to know

Morning.  We are dealing with an AI Grievance at the moment, apart from the time its taking (which so far has been 3 times the amount it usually does), the employee in question has really struggled when in meetings.  They've relied so much on AI that what they say and what is written does not really match up.  They are unaware that they need to talk us through the situation in detail and instead have relied on their letter.  They brought notes and even read out "do not say this part, its for your notes only".  We actually pointed that out to them and tried to start again.  Questioning slowly, asking them to go back to the beginning and talk to us in their own voice, with their own opinions.  

Due to the issues this is causing we are looking at our policies and are going to add some information regarding the use of AI.  Organise training for managers (having that difficult conversation easier) and we've even used AI ourselves to see if it can point out the main issues in received grievances.  For example, 1 grievance was 15 pages long... AI cut it down to 3 issues that needed to be discussed.  

Hi Jo, thank you for sharing this experience. It sounds like this area, that always could pose difficulties, is now posing even more challenges to process and resourcing. Thank you for the solutions you have suggested as well.

A couple of practical responses to add to your list:

1. Put accountability in the grievance policy itself. One line saying the employee owns every statement they submit, however it was drafted, and will be asked to talk it through at the first meeting. That makes the early human conversation the default rather than a favour, and it handles the "can't explain their own letter" problem Jo described.
2. Add a data line to the AI use policy. An AI-drafted grievance often means colleagues' names, health details or internal emails went into a public chatbot. Say plainly what must never go into an unapproved tool, and apply the same rule to HR: if you use AI to cut a 15-page grievance down to the core issues, use an approved tool and check the summary against the original.

For managers, a short session built around one anonymised example tends to land better than a long module: find the core issues, ask clarifying questions, and don't dismiss it as "just AI".

Drafted with AI assistance and reviewed by me.

Hi Holly, building on Fran's point about the early conversation: the step I've found most useful is to end that conversation with a short written summary, agreed with the employee, of what the grievance is about and what outcome they want. One or two sentences. The full submission is kept and read, but the investigation is scoped to that summary.

It helps on two of your challenges at once. It shortens the time spent working out what a long submission is actually about, and it gives the employee a clear chance to put the issue in their own words before anything is formalised.

One addition to your response list: train managers to separate the format of a grievance from its substance. A long, legalistic submission isn't automatically serious, and a short, informal one isn't automatically minor.

A splendid and amusing critique of the pitfalls of using AI.

Thank you for these suggestions Fran, they seem very sensible and practical! Do let me know how this lands