12 ChatGPT prompts for HR that actually work (with the ones that don't).
Most HR teams are using AI as a fancy spell-checker. Here are the prompts that compress 2 hours of work into 10 minutes — and the ones that look impressive in a demo but break in production.
PJ
Pawan Joshi
Global HR & Operations
Share
Two years into the AI-in-HR era, here is what I keep seeing: 80% of teams are using ChatGPT or Claude as a fancier grammar checker. Maybe 15% are using it for first drafts. Only about 5% are using it for the things that actually move the needle — and those 5% are running circles around everyone else.
Below are the prompts I keep returning to. They are organized by what they replace, not by what they sound impressive doing. I have included three categories I tell people to avoid — they look good in a demo and quietly cause damage when scaled.
Where AI is actually creating value in HR right now
Survey of 240 HR leaders using LLM tools daily, 2025.
73%
report time saved on JD writing & rewrites
SHRM AI Adoption, 2025
58%
report higher quality first drafts of comms & policies
SHRM AI Adoption, 2025
11%
report measurable improvement in hiring outcomes
SHRM AI Adoption, 2025
2x
faster turnaround on routine HR documents
Mercer Pulse, 2025
The 12 prompts that actually work
1. Turn a half-baked role idea into a structured JD
2. Audit an existing JD for bias and noise
3. Generate interview questions from a JD
4. Rewrite a rejection email so it doesn't burn the brand
5. Summarise an employee survey into actionable themes
6. Stress-test a policy before you ship it
7. Draft a performance improvement conversation script
8. Translate a complex policy into plain language
9. Turn 1:1 notes into a development plan
10. Pre-brief a candidate debrief
11. Build a manager training session from scratch
12. Audit your own work week
The 3 prompts to avoid
What to skip
Avoid these
'Write me a performance review for [name].' — flattens nuance, sounds robotic, hurts trust if discovered.
'Screen these resumes and rank them.' — bakes in bias, can't be audited, legally risky in EU & NYC.
'Predict who is likely to quit in the next 6 months.' — high false-positive rate, damages culture if leaked.
Do this instead
Use AI to draft the structure, write the review yourself — your judgement is the product.
Use AI to write smarter application questions; let humans do shortlisting.
Use AI to summarise themes from engagement data — let managers act on signal, not predictions.
How to know it's actually working
Pick one workflow this week. Time how long it currently takes (JD writing, survey synthesis, 1:1 prep, whatever). Run it with one of the prompts above for two weeks. Measure both time saved AND quality (let a peer rate it blind). If both don't improve, the prompt isn't right for your context — iterate or drop it. Most teams over-invest in tools and under-invest in this measurement loop.
“AI in HR is a magnifier, not a multiplier. Give it good judgement and it will move mountains. Give it lazy thinking and it will scale your laziness faster than you can apologize for it.”
Found this useful? Share it.
Written by
Pawan Joshi
HR & Operations leader scaling global remote teams across Nepal, the Philippines, Australia, and the US. Tech-leaning writing lives on Medium.
