Best ChatGPT Prompts for Work: 40 That Earn Their Keep

Most "best ChatGPT prompts" lists are a parade of one-liners that work in demos and fall apart in real meetings. The 40 prompts below are different. They have been used by working professionals — managers, consultants, marketers, recruiters, finance leads — for at least one quarter and survived contact with reality. Each is a structured prompt, not a magic phrase, and each is paired with the kind of role and audience it expects you to fill in. The categories below match the work most knowledge workers actually do all day: writing email, prepping meetings, giving feedback, planning strategy, working in spreadsheets, writing documentation, hiring, and having difficult conversations.

Table of contents

• Email drafts
• Meeting prep
• Performance feedback
• Strategy and planning
• Spreadsheet logic
• Documentation
• Hiring and interviews
• Difficult conversations
• Frequently asked questions
• The bottom line

Email drafts

The email category is where most professionals first feel ChatGPT pay back. Every prompt below assumes you paste in the original thread or context first.

1. The reply rewriter. "Rewrite my reply below to sound warmer but still firm. Keep my voice — direct, light contractions, no corporate jargon. Cut anything that does not earn its place. Two paragraphs, max."

2. The unsubscribe-but-keep-the-relationship. "Draft a reply that politely declines this proposal in a way that does not burn the bridge. Acknowledge the work they put into the deck. Do not promise future contact unless I would actually mean it."

3. The chase that does not nag. "Write a follow-up to a contact who has not replied to my email of [date]. The tone should be relaxed and assume goodwill, not impatience. Mention something specific from the original thread to refresh their memory."

4. The one-line ask, expanded. "Here is the rough one-line version of what I want to ask. Expand it into a polite, professional email of about 120 words. Make it easy for the recipient to reply with a yes or no."

5. The bad-news send. "We have to tell a customer their integration request will not be in this quarter's roadmap. Write the email. Acknowledge the impact on them, do not over-apologise, and offer one concrete next step."

Meeting prep

The 30 minutes before a meeting are where ChatGPT earns the most leverage of any work category, because the alternative is usually no prep at all.

6. The agenda builder. "I have a 45-minute meeting with [role] about [topic]. The desired outcome is [decision]. Build me a tight agenda with time boxes, the questions I should ask, and the order to ask them in to make the desired outcome most likely."

7. The pre-read summariser. Paste in a long deck or document. "Summarise this in five bullets for someone walking into a meeting in 15 minutes. Flag the two assumptions in the deck most worth questioning."

8. The objection rehearsal. "I am presenting [proposal] to [audience]. List the five strongest objections they will raise, ranked by how hard they are to answer. For each, give me a one-paragraph response that does not sound defensive."

9. The agenda preview email. "Convert the agenda above into a 90-word note I can send to attendees the night before, framing the desired outcome and what I need from them."

10. The meeting recap. Paste the meeting transcript or your notes. "Produce three things: a one-paragraph summary, a numbered list of decisions made, and a numbered list of action items with owners and dates."

Performance feedback

Performance feedback is the workplace use that benefits most from a critic looking over your shoulder. Vague feedback is universal. Vague feedback also damages people. The prompts below force precision.

11. The behaviour translator. "Here is some feedback I want to give: '[your draft].' Help me tighten it by replacing any vague adjectives ('proactive,' 'collaborative') with specific observed behaviour and impact. Keep it under 200 words."

12. The Situation-Behaviour-Impact reformatter. "Convert this feedback into the SBI format. Make sure the situation is concrete, the behaviour is observable, and the impact is named in business terms."

13. The strengths balance check. "Read this draft of an annual review. Tell me whether the strengths and growth-areas feel proportionate, given the rating I am giving. Highlight anything that contradicts the rating."

14. The hard message rehearsal. "I need to tell [person] that their work this quarter was below the bar. Help me draft what I would say in a 1-1, in two paragraphs of plain spoken language. Anticipate their pushback and write what I would say in response."

15. The peer-review distiller. Paste anonymous peer feedback. "Synthesise the consistent themes across these comments into three strengths and three growth areas. Quote the feedback that most clearly supports each."

Strategy and planning

16. The first-principles strategy critique. Paste a draft strategy doc. "Critique this strategy from a first-principles perspective. Where are the assumptions doing the heavy lifting? What would have to be true for this strategy to fail? What is the cheapest test of those assumptions?"

17. The red-team review. "You are a sceptical board member who has seen this kind of plan fail twice. Read the doc and ask the five questions you would ask in the meeting that would put the team on the back foot."

18. The OKR sanity check. "Here are our team's draft Q3 OKRs. Tell me which key results are actually inputs masquerading as outcomes. Suggest a true-outcome version for any you flag."

19. The competitor reality check. "We claim our differentiator is [X]. List five reasons a customer might not see it as a differentiator. For each, write the rebuttal. Then tell me which rebuttals you find weak."

20. The trade-off forcer. "We are debating between [option A] and [option B]. Build me a trade-off matrix with the five dimensions that matter most for this decision. Score each option from a sceptical outsider's view, with reasoning."

Spreadsheet logic

For Excel and Sheets work, the model is closer to a junior analyst than a senior one. Trust formulas it produces, but always run the numbers on a small sample before scaling.

21. The formula explainer. Paste a complicated formula. "Explain what this formula does, line by line. Tell me if there is a simpler way."

22. The formula writer. "I have data with [columns]. I want to compute [outcome]. Write me the Excel formula and explain why it works."

23. The pivot logic check. "I want a pivot showing [rows] vs [columns] aggregating [measure]. Tell me what filters and groupings I need, and what gotchas to watch for."

24. The Power Query trick. "I have this CSV with [problem]. Walk me through the Power Query steps to clean it. Be specific."

25. The model audit. Paste the formulas in a financial model. "Audit this for circular references, hard-coded values that should be assumptions, and any formula a sceptical reviewer would flag."

Documentation

26. The release-note generator. Paste raw commit messages. "Convert these into customer-facing release notes. Group by feature, bug fix, and removed. Use the voice of someone who actually uses the product."

27. The PRD critic. "Critique this product requirements doc. List ambiguities engineering will hit during build. List acceptance criteria that are non-falsifiable."

Hiring and interviews

28. The job description rewriter. "Rewrite this job description to remove cliches, add specifics about scope and seniority, and make it appealing to the strong candidates we want, not the volume we will get."

29. The screening question generator. "Build me five 30-second screening questions that will distinguish a senior [role] from a mid-level. The questions should not be Googleable."

30. The competency-matrix writer. "Write a competency matrix for [role] with four levels (junior, mid, senior, staff). For each level, list five observable behaviours."

31. The interview-debrief structurer. "Convert these unstructured interview notes into a hire/no-hire memo. Cite specific evidence for each scoring dimension. Flag any bias risk."

32. The reference call cheat sheet. "I am about to call a reference for [candidate]. Build me five questions that will surface real signal, not just polite confirmation."

Difficult conversations

33. The conflict de-escalator. "I have to talk to [colleague] about [recent friction]. Help me draft an opener that names the issue without blame and invites a real exchange."

34. The boundary setter. "[Person] keeps asking me to do work that is outside my role. Help me write a clear, kind decline that protects my time without damaging the relationship."

35. The exit conversation. "I am giving notice. Help me draft what I will say to my manager — appreciative, professional, and not over-explanatory."

36. The bad-fit hiring conversation. "I need to tell a candidate they have not progressed past round two. Write a note that is specific enough to be respectful and short enough to send fast."

37. The pay-rise conversation prep. "Help me prepare for a pay-rise conversation. Build me the case structure: market, my last 12 months of impact, and the ask. Anticipate three pushback patterns and how I would respond."

38. The board-update bad-news section. "Help me draft the section of the board update where we report missing the quarter. Acknowledge, attribute, and lay out the corrective plan in plain language."

39. The vendor escalation. "Draft a firm but professional escalation to [vendor]'s account manager. Include the specific failures by date, the contractual obligation breached, and the resolution we expect."

40. The team-restructure announcement. "Help me draft the all-hands message announcing this team restructure. Be honest about what is changing, name the people impacted, and do not soften the parts that should not be softened."

The pattern across all 40 is the same: explicit role, explicit context, explicit format, explicit constraint. None of these prompts will work as a one-line copy-paste. Each is a frame you fill with specifics. For a deeper look at why these structures work, see our prompt engineering hub. For broader context on ChatGPT for office work, the pillar guide is the next stop.

Frequently asked questions

Should I share these prompts with my team?

Yes. The biggest unlock most teams find is not better individual prompting — it is shared prompt libraries. A team of five with a shared bank of 40 vetted prompts will outperform a team of five each maintaining their own private list. Custom GPTs are a natural way to operationalise the best ones.

Will my employer's policy let me use these?

Most policies in 2026 allow ChatGPT for general work as long as you do not paste confidential or regulated data into a personal account. If your company has Team or Enterprise, use those — they do not train on your data and the audit trail is cleaner. If they do not, ask. Most policy answers are simpler than people fear.

Can I trust the prose ChatGPT produces for important writing?

For external communication, treat it as a strong first draft, not a final draft. The model has measurable tells — overused transitions, slightly elevated diction, predictable rhythm — that experienced readers pick up on. For internal work, especially under deadline, the cost of polish you would not have done anyway is low. For external writing, edit for your own voice.

What about the privacy of these prompts?

Anything you paste into Free or Plus may be used to train future models unless you turn the setting off. For genuinely sensitive content — performance reviews, internal restructures, financials — do not use a personal Free account. Either turn off training in settings or use Team or Enterprise.

How do I know if a prompt is good?

Three tests. Did it produce something you would have spent more than ten minutes producing yourself? Did the answer require fewer than three iterations? Could a colleague pick up the same prompt cold and get a comparable result? If yes to all three, the prompt is earning its place.

The bottom line

The 40 prompts above are a starting library, not a finished one. Treat them as templates. Iterate on the ones that almost work. Delete the ones that never do. The professionals getting the most out of ChatGPT in 2026 are not the ones with the cleverest single prompt — they are the ones who maintain a small, ruthless library of prompts that have survived three months of real use. Start with five. Use them every day. Add a sixth only when you find one worth keeping. The rest of the ChatGPT hub covers the platform features that turn a prompt library into a team capability.

Last updated: May 2026