← All tools
Diagnostic tool
What can generative AI touch?
Generative AI is not one risk category. The same assistant is unremarkable for one task and unacceptable for another — and most organizations approve it once, then never reassess when the use changes. Sort eight everyday HR tasks.
The three categories
Safe
Safe with ordinary care
A human owns the output, no individual is scored, and a mistake is visible and correctable.
Controls
Needs controls before you do this
Legitimate, but it touches a person or their data — so it needs documentation, oversight and notice first.
Do not
Do not use a general model for this
Either the consequence is too direct, or the data should not enter a general-purpose model at all.
Task 01
Drafting a job advertisement
Recruiter provides the requirements, edits the draft, and posts it.
Task 02
Summarizing a shortlisted CV for the hiring manager
Recruiter pastes in the CV and forwards the summary.
Task 03
Asking which of two candidates is the better fit
Recruiter is undecided and asks the assistant to weigh them up.
Task 04
Answering an employee question about leave policy
Assistant is grounded in the employee handbook and cites the section.
Task 05
Writing a performance review from a manager's notes
Manager provides bullet points, the assistant produces the narrative.
Task 06
Summarizing an employee's medical or accommodation file
HR pastes documentation to produce a summary for a case meeting.
Task 07
Translating an HR policy into another language
Existing published policy, translated for a regional office.
Task 08
Drafting talking points for a retention conversation
Manager provides context about an employee considering leaving.
Indicative guidance only, based on your answers. Not a legal determination, a conformity assessment, or a substitute for advice from qualified counsel in the relevant jurisdiction. Regulatory requirements, effective dates, and enforcement practice change; verify current obligations before acting. Nothing you enter here is transmitted or stored.