10 AI Prompts for HR Leaders That Save 6 Hours a Week

Every other HR leader in 2026 says "we use AI." Asked "how exactly," 80% answer "well, through ChatGPT for emails." That's not AI in HR. That's AI as a fancy word processor.

Real savings in time and quality in HR work through AI aren't magic and aren't "add a chatbot to the intranet." They're a set of concrete prompts for recurring tasks: meeting prep, framing feedback, analyzing survey results, drafting a job description. Each prompt is a template you set up once for yourself, then use dozens of times.

This article is ten such prompts with templates, examples, and instructions. You can copy them, you can adapt them. They've all been tested in our customers' HR teams over 2025–2026.

An important note on boundaries: AI does the preparatory work, not the final texts. All 10 prompts are for drafts you edit before sending. If you want "AI will write it for me and send it," this isn't that article. Here it's about saving hours without losing control over the content.

Why Frame Prompts Thoughtfully

The main reason AI in HR "doesn't work" isn't a bad AI, it's a bad prompt. "Write feedback for Anna" isn't a prompt, it's an instruction to the model to guess. It'll guess poorly, the answer will be generic, and you'll say "AI is useless."

A good prompt is four elements, always together.

Task. What I want as output. Not "help me," but "draft a thank-you."

Context. What the model needs to know about the situation. Who it's for, what happened, any tone constraints, the relationships on the team.

Constraints. What to do and not do. "No formal phrasing," "don't mention salary," "avoid abstractions."

Output format. What form the answer takes. "3 options to choose from," "a list of 5 points," "2–3 sentences for chat."

All four elements are mandatory. Skip one and the model guesses, and quality drops. With all four, the result becomes far better, and you usually only need to edit it by 10–20%, not rewrite from scratch.

Now — ten concrete prompts, each with a template and an example.

Ten Prompts in Four Groups

I'll go through each of the ten. For each: when to use it, the prompt template, what to fill in, and what to do with the answer.

Group 1. Preparation

Prompt 1. A 1:1 brief.

When to use: an hour before the meeting, to refresh context.

Template:

I'm preparing for a 1:1 with a direct report. Here are notes from the last 3 meetings: [paste notes]. Here are observations from the interval between meetings: [paste activity and signals]. Give me a short 4-point brief: what we discussed last time, what was left unanswered, what changed in the interval, and 2 recommended questions to ask. No boilerplate, be specific.

What to do with the answer: read it, check that the model didn't invent facts (hallucinations are a real problem), adapt the questions to your style.

Prompt 2. Structuring observations for a performance review.

When to use: at the start of preparing a review, when you have "a pile of material" but no structure.

Template:

I'm preparing a performance review for an employee. Here are my observations for the period: [paste notes, peer feedback, project numbers]. Group them into this structure: 1) context of the role and projects, 2) what went well — 2–3 concrete situations with behavior and result, 3) what didn't go well — 1–2 situations with the cause, 4) a recommended next step. Only what's in my observations. Don't add anything I didn't write.

What to do with the answer: check that the structure fits, rewrite it in your own words (don't copy). This is a base, not a final document.

Prompt 3. A short summary of a long meeting.

When to use: after a two-hour strategy meeting or a big retrospective.

Template:

Here are notes from a meeting [topic, duration]: [paste notes or transcript]. Make a summary in this format: 1) key themes (3–5 points), 2) decisions made with owners, 3) open questions requiring a next discussion. No more than 200 words.

What to do with the answer: check against your memory that the AI correctly understood who said what. Publish after a manual review.

Group 2. Communication

Prompt 4. A draft thank-you with specifics.

When to use: when you want to praise someone, but "thanks for the work" sounds formal.

Template:

I want to publicly thank a colleague. Context: [name, role, what exactly they did and why it matters]. Give 3 versions of a thank-you: one formal for the company feed, one warm for the team chat, one very short for a direct message. Specifics required, no "great job" or "well done."

What to do with the answer: pick a version, edit it to your tone, send it. Don't copy verbatim — the honesty gets lost.

Prompt 5. Rephrasing harsh feedback.

When to use: when you've written a feedback draft and realize it's too harsh or, conversely, too vague.

Template:

I wrote feedback for a direct report, but it came out [harsh / vague / abstract]. Here's the text: [paste]. Rephrase it as a concrete description of behavior and its consequence, without evaluative words about the person themselves. Keep the meaning and don't add anything new.

What to do with the answer: read it aloud, imagine this feedback being said to you. If it sounds fine, send it. If it's still too much, ask for another rewrite.

Prompt 6. A message about difficult news to the team.

When to use: when you need to inform the team about a reorg, a layoff, a canceled project.

Template:

I need to inform a team of [N people] about [difficult news]. Context: [what happened, why, what changes]. Draft a message to the team that: 1) states the main point right away, 2) explains the reasons briefly and honestly, 3) describes what specifically changes, 4) gives a contact for questions. No euphemisms or formalities. Tone — serious but human.

What to do with the answer: definitely rewrite the key sentences in your own words. AI structures well, but a "human voice" in a crisis message can't be faked. Run it by someone you trust before sending.

Group 3. Analytics

Prompt 7. Finding themes in open survey answers.

When to use: after a survey with open-ended answers (even a single "comment" field) — to quickly see the themes.

Template:

Here are open-ended answers from an anonymous employee survey: [paste answers]. Find 5–7 common themes that appear across multiple answers. For each theme: 1) a short name, 2) roughly how many answers touch on it, 3) an example phrasing (without identifying the author). Don't draw conclusions, just describe the themes.

What to do with the answer: double-check by hand for at least 2–3 themes (pull the answers and confirm the AI didn't merge incompatible ones). Then use it as a starting point for your own analysis, not as a finished report.

Prompt 8. Hypotheses for turnover causes by segment.

When to use: when you see turnover in one department above baseline and need to form hypotheses.

Template:

In the [name, size, specifics] department over the last 6 months, [N] people left. Reasons for leaving (from exit interviews): [paste aggregated reasons]. Additional context: [important events — reorg, new manager, changed goals]. Formulate 4–5 hypotheses about the structural causes of turnover in this department. Not "maybe pay," but concrete hypotheses you can test through a 1:1 or a measurement.

What to do with the answer: pick 1–2 hypotheses that resonate and test them empirically. Don't take it as an answer — it's a starting point for a conversation.

Group 4. Content

Prompt 9. A job description from a set of requirements.

When to use: when you have a set of role requirements but no coherent text.

Template:

I'm opening a role: [title, level]. Team: [context]. Responsibilities: [list]. Requirements: [list]. What we offer: [list]. Write a job posting where: 1) the first paragraph is about the team and the problem the role solves (not the company in general), 2) responsibilities are framed as 5–7 concrete results in 6 months, 3) requirements are split into "must-have" and "nice-to-have," 4) what we offer is concrete, no clichés. No "dynamic, fast-growing company" or "friendly team."

What to do with the answer: edit it for your company, check that the posting has something that distinguishes your position from the typical template. If not, ask for a rewrite emphasizing the differences.

Prompt 10. An onboarding checklist for a specific role.

When to use: when hiring a new role (or an existing one into a new department), to quickly assemble a first-week plan.

Template:

I hired a new employee for the role of [title, team context]. Make a detailed first-week plan day by day: what to do, who to introduce them to, what tasks to give. For each day — 3–5 points. Note that this is a [role] whose specifics are [context]. The goal — by the end of the first Friday, the person can answer 7 basic questions: who's their manager, who's the team, what to do in the first month, which tools, where to find answers, when the first 1:1, how we do things here.

What to do with the answer: walk through the plan, add your company's specifics (buddy names, concrete documents, real links). AI gives the scaffold — you fill in the details.

What NOT to Do With Prompts in HR

A few rules we treat as mandatory on our team. Breaking them undermines all the rest of your work with AI.

Don't send sensitive data to public models. Salaries, evaluations, medical notes, specific complaints with names — you can't throw these into a public AI that has no contract with your company. Use a corporate AI with a DPA, or anonymize the data before sending. This isn't paranoia — it's a data-protection requirement.

Don't copy the answer verbatim. Especially for emails and feedback. AI writes in a recognizably "AI-ish" way, and after 2–3 times employees start noticing. Trust drops sharply. A draft — yes; sending without editing — no.

Don't use AI for final decisions about people. Who to promote, who to let go, who to give the hard project — that's a decision for the manager and HR partner. AI can help gather material, but the decision stays with the human. This is both about ethics and about the fact that AI errs more often than it seems.

Don't confuse "got an answer" with "got the truth." AI speaks confidently even when it's wrong — this is called hallucination. It's especially dangerous in text analysis: the model can "see" themes that aren't in the data, or attribute an answer to the wrong respondent. Always verify by hand for important conclusions.

Don't use the same prompt without adapting it. The templates in this article are a starting point. For your company, culture, and specifics, fine-tuning is needed. One prompt works on one team and doesn't on another. There are no "universal prompts for all time."

The Bottom Line

AI saves an HR leader real hours — but only if you use it thoughtfully, not "through ChatGPT for emails." The ten prompts in this article cover most of the recurring work: prep for meetings and reviews, framing feedback and messages, analyzing open-ended answers, creating content.

The main principle: four elements in every prompt — task, context, constraints, format. Without any of them, the model guesses, and the result disappoints.

And the main boundary: AI makes drafts, the human makes the final texts and decisions. We hold this line firmly. Without it, AI becomes not an assistant but a source of risk.

In the next issue we'll go through how to sell a recognition program to a CEO on a single page — twelve concrete numbers that work as an argument in most companies. It's material for pitching an idea inside the company, but it's equally useful for HR leaders defending a budget.

If you'd like to receive a weekly digest of tested HR prompts (with examples and commentary on the answers), subscribe at the link in the card. Free, unsubscribe in one click.