Last month, a friend who runs people ops at a 90-person SaaS company asked me to help her "get better at AI." What she actually meant was: stop spending three hours writing job descriptions that read like legal disclaimers, and stop drafting the same performance review comments she wrote last year, and the year before that.
We didn't spend a week learning prompt engineering theory. We built a set of prompts — specific, ready-to-paste ones — for every part of her job. By Friday she'd cut her admin writing time by more than half. Not because AI is magic, but because she stopped starting from blank every time.
These are the prompts. All 60 of them.
Why HR is one of the best AI use cases nobody talks about
HR work is document-heavy, repetitive, and formula-dependent in ways that most roles aren't. A recruiter writes hundreds of job postings. A people ops manager sends the same types of performance feedback, onboarding emails, and policy announcements dozens of times a year. The underlying structure barely changes — the details do.
That's exactly the gap AI fills well. Give it a strong template, add the specifics, and you get a solid first draft in 30 seconds instead of 30 minutes.
The prompts that work best in HR share three traits: they're specific enough that the AI has something real to work with, they include fill-in brackets for the parts that change each time, and they tell the AI what not to do as much as what to do (no jargon, no vague corporate-speak, no passive voice).
The six categories worth your time
1. Job postings and recruiting copy
Most job descriptions are terrible. They list 14 responsibilities, require 7+ years for a mid-level role, and somehow still manage to sound generic. AI can fix this — but only if you tell it what you actually want.
Prompt that works:
Write a job posting for a [JOB TITLE] at [COMPANY NAME]. We're a [INDUSTRY] company with [COMPANY SIZE] employees. Our culture is [DESCRIBE CULTURE]. Include: role summary, 5-7 key responsibilities, required qualifications, preferred qualifications, and a benefits section. Tone: [FORMAL/CONVERSATIONAL/ENERGETIC].
The key detail here is the tone instruction. Without it, AI defaults to whatever "average job posting" sounds like, which is exactly what you don't want.
2. Interview questions and evaluation
Generic interview questions get generic answers. "Tell me about a time you showed leadership" doesn't filter for much. The prompts that work here push AI toward behavioural and situational formats tied to specific competencies.
Prompt that works:
Generate a structured interview guide for a [JOB TITLE] role. Include: 5 competency-based questions with follow-up probes, 3 situational questions, 2 culture-fit questions, and a scoring rubric (1-4) for each. Competencies to assess: [LIST COMPETENCIES].
The scoring rubric request is the move. It forces AI to define what a "good" answer looks like, which saves you from having to calibrate across a panel of interviewers ad hoc.
3. Performance reviews
Performance review season is when HR professionals spend days writing variations of the same five sentences. AI handles this well because the format is consistent — you're always writing about what someone did, what the impact was, and what they should do differently.
Prompt that works:
Turn these bullet-point notes into a professional, balanced performance review for [EMPLOYEE NAME] in the [JOB TITLE] role. Notes: [PASTE NOTES]. Tone: direct, constructive, specific — not vague or over-hedged.
The "not vague or over-hedged" instruction matters. AI left to its own devices will write the softest possible version of every piece of feedback. You have to explicitly push back on that tendency.
4. Onboarding materials
New hire documents age quickly and are rarely updated. The problem is that updating them takes just as long as writing them from scratch — or feels like it does. AI makes this fast enough that there's no excuse to keep sending outdated welcome guides.
Prompt that works:
Write a 'What Great Looks Like' document for a new [JOB TITLE]. Describe what outstanding performance looks like at 30, 60, and 90 days. Be specific — avoid vague phrases like 'hit the ground running'.
"Avoid vague phrases like 'hit the ground running'" does a lot of work. Naming the specific clichés you want to avoid trains the output to be concrete.
5. Employee communications and policy
This is where HR writing gets the most generic. Policy announcements, all-hands agendas, return-to-office memos — they all start sounding the same. The fix is to give AI your actual company voice and specific context, not just a topic.
Prompt that works:
Write a company-wide announcement for [NEWS/CHANGE]. Tone: transparent, direct, and human. Include: what's changing, why, timeline, and who to contact with questions.
The word "human" in the tone instruction pulls AI away from its corporate-memo default. Worth including in almost every comms prompt.
6. Training and L&D
Designing a training session from scratch takes hours. Designing a good one takes even longer, because it requires thinking through learning objectives, activities, and how to measure whether anything actually landed. AI can get you 80% of the way there, fast.
Prompt that works:
Design a 60-minute manager training session on [TOPIC]. Include: learning objectives, agenda with timings, a group activity, key takeaways, and post-session action items.
What makes a prompt actually work
Three things:
Specificity beats generality every time. "Write a job posting" gives you mush. "Write a job posting for a customer success manager at a 50-person B2B SaaS company with a remote-first culture, conversational tone, and a real emphasis on problem-solving over ticket volume" gives you something you can actually use.
Tell it what tone to avoid. "Not a LinkedIn post" and "not corporate-fluffy" and "not vague" are legitimate prompt instructions. AI defaults to the average of whatever it was trained on. Explicitly naming what you don't want is how you get out of that average.
Paste your existing documents in. If you have a job description that worked, paste it into the prompt as a reference. If you have a performance review format, include it. AI's output improves dramatically when it has something real to work with, not just an abstract request.
I put all 60 prompts — across all six categories — into a PDF you can keep open in a second tab while you work. It's at ghostweasel.gumroad.com/l/supdc for $9. Every prompt has fill-in brackets so you can paste and customise in under a minute.
If you're spending more than 20 minutes writing any HR document that exists in some form at every company, you're leaving time on the table. These prompts don't automate the judgement part of HR — they just handle the typing.
Top comments (0)