ChatGPT prompts for work: 42 copy-paste templates by role

The best ChatGPT prompts for work give the model a role, the context, a clear task and the format you want back, and they leave room for ChatGPT to ask questions. Below are 42 copy-paste templates grouped by role, plus how to adapt them, save them for reuse, and keep company data safe.

What makes a good ChatGPT prompt for work?

A good ChatGPT prompt for work tells the model who it is, what situation it is in, what to produce, in what shape, and within what limits. Most disappointing answers come from prompts that skip the context, not from the model.

OpenAI's own prompt engineering best practices boil down to three habits: be clear and specific, refine iteratively, and name the tone you want. The API prompting guide adds two more that help in ChatGPT too: give examples of the output you want, and separate instructions from source material with clear markers.

Put together, a work prompt has five parts:

Here is the same request written badly and well:

Write an email to a client about the delay.

You are a project manager at a design agency. Our client, a dentalclinic, expected their new website on March 3. It will launch March 17because their content arrived two weeks late. Write an email to theclinic owner that explains the new date, does not blame them, and asksfor final photos by March 10. Under 150 words, warm and professional.

The second version takes 30 more seconds to write and saves several rounds of "make it shorter, less formal, mention the photos."

How do you adapt these templates?

Replace every [bracketed placeholder] with real specifics, paste the source material below the prompt, and ask follow-up questions instead of starting over. The templates below are starting points; the details you add are what make the output usable.

Three techniques make any template work better:

1. Paste the material. Put the email thread, notes or data under the prompt, separated with a line like --- or wrapped in quotes, so the model knows what is instruction and what is input.
2. Ask for questions first. For anything new or high stakes, add: "Before you write, ask me up to five questions about anything that would change your answer."
3. Iterate in the same chat. Follow-ups such as "shorter," "more direct," "now as a table" or "what would a skeptical reader object to?" are faster than rewriting the prompt.

If you want the theory behind these habits, our prompt engineering guide covers examples, structure and iteration in more depth.

Prompts for email, meetings and everyday writing

These prompts cover the communication work everyone does: replying, summarizing, preparing and editing. They work for any role.

Email

1. Reply to a difficult email:

Here is an email I received. Draft a reply that [goal, e.g. declinesthe request but offers an alternative]. Keep my relationship with thesender in mind: [relationship]. Under [N] words. Tone: [tone].---[paste email]

Why it works: stating the goal and the relationship stops ChatGPT from defaulting to a stiff, overly apologetic reply.

2. Turn notes into a clear email:

Turn these rough notes into an email to [recipient and role]. Lead withthe decision or ask in the first sentence, then the reasons in bullets.End with one clear next step and a date.---[paste notes]

3. Summarize a long thread:

Summarize this email thread for someone who has not read it. Give:the current status in one sentence, decisions made, open questions,and who owes what by when. Use a short table for action items.---[paste thread]

Why it works: a fixed structure (status, decisions, questions, owners) makes the summary scannable and exposes gaps in the thread itself.

4. Shorten anything:

Cut this message to half its length without losing any request ordeadline. Keep my voice. Show the shortened version only.---[paste text]

Meetings

5. Prepare an agenda:

I am running a [length]-minute meeting with [attendees and roles] todecide [decision]. Draft an agenda with timings, the pre-read eachperson needs, and the question we must answer by the end.

6. Turn a transcript into minutes:

From this meeting transcript, write minutes with: decisions, actionitems (owner, task, due date), open risks and unresolved questions.Do not invent owners or dates; write "unassigned" if none was given.---[paste transcript]

Why it works: the explicit rule against inventing owners and dates prevents the most common error in AI meeting notes. For tools that record and summarize meetings automatically, see our guide to AI meeting notes.

7. Prepare for a tough conversation:

I need to talk to [person/role] about [topic]. Here is the background:[context]. List the three objections they are most likely to raise anda calm, respectful response to each. Then suggest an opening line.

Editing and thinking

8. Edit for clarity:

Edit this for clarity and concision for [audience]. Keep technicalterms that [audience] knows. Mark each change in bold and list thethree biggest problems you fixed.---[paste text]

9. Stress-test a plan:

Act as a skeptical [CFO / senior engineer / customer]. Read this planand list the five weakest points, ordered by risk. For each, say whatevidence would change your mind.---[paste plan]

Why it works: assigning a skeptical persona gets you criticism instead of polite agreement, and asking for disconfirming evidence turns objections into a to-do list.

10. Explain something outside your field:

Explain [concept] to me as a [your role] with no background in[field]. Use one concrete example from [your industry]. Then give methree questions I could ask an expert to check my understanding.

For more on getting a human tone out of drafts, see how to make AI writing sound human.

Prompts for managers and team leads

Managers get the most value from ChatGPT on preparation and structure: one-on-ones, feedback, plans and updates. Never paste performance details that identify someone unless your company allows it.

11. One-on-one preparation:

I have a 1:1 with a [role] who [situation, e.g. seems disengaged sincea reorg]. Suggest five open questions that invite candid answers, andtwo things I should avoid saying.

12. Write constructive feedback:

Help me write feedback using the situation, behavior, impact format.Situation: [when/where]. Behavior: [what they did]. Impact: [effect].I want them to [desired change]. Keep it specific and kind, under 120 words.

Why it works: the situation, behavior, impact format keeps feedback about observable actions rather than personality, which makes it easier to hear.

13. Weekly status update:

Turn these bullet notes into a weekly update for [audience]. Sections:progress, risks (with mitigation), decisions needed from them. Leadwith the most important item. Under 200 words.---[paste notes]

14. Plan a quarter:

Our team of [N] people owns [area]. Company goals this quarter: [goals].Propose three team objectives with two measurable key results each,and list what we should stop doing to make room.

Why it works: asking what to stop doing forces trade-offs, which is the part of planning that teams most often skip.

15. Delegation brief:

Write a delegation brief for [task] that I am handing to [person,experience level]. Include: goal, why it matters, what done looks like,decisions they can make alone, and when to check in with me.

Prompts for sales, marketing and customer support

Customer-facing teams use ChatGPT for research, first drafts and consistent replies. Always check facts about your product and pricing yourself; the model only knows what you paste in.

Sales

16. Account research brief:

I am meeting [company] next week. From the notes below, write aone-page brief: what they do, likely priorities for a [buyer role],three questions to ask, and how [our product] might fit.---[paste website text, LinkedIn notes, news]

17. Cold email:

Write a cold email to a [role] at a [company type] about [problem wesolve]. One specific observation about their business: [observation].Under 90 words, no buzzwords, end with a low-effort question.

Why it works: one real observation about the prospect does more than any clever phrasing, and the word limit keeps the email readable on a phone.

18. Objection handling:

Our product is [description], priced at [price]. List the eight mostcommon objections a [buyer role] would raise and a short, non-pushyresponse to each, grounded in these facts: [facts].

Why it works: grounding responses in facts you supply keeps ChatGPT from promising features or terms your product does not have.

19. Call follow-up:

From these call notes, write a follow-up email that recaps what theprospect said in their words, confirms next steps and dates, andattaches nothing we did not promise.---[paste notes]

Marketing

20. Content brief:

Create a content brief for an article targeting "[keyword]" for[audience]. Include: search intent, the questions the article mustanswer, a suggested outline with H2s, and what would make it betterthan a generic article.

21. Ad copy variations:

Write 10 headline and description pairs for a [platform] ad for[product]. Audience: [audience]. Main benefit: [benefit]. Respectthese character limits: headline [N], description [N]. Vary the angle:price, outcome, fear of missing out, social proof, curiosity.

Why it works: naming the angles produces variety you can test, instead of ten versions of the same headline.

22. Repurpose one piece of content:

Turn this article into: one LinkedIn post (under 200 words), threeshort social posts, and a 5-bullet newsletter summary. Keep claimsexactly as they appear in the article.---[paste article]

23. Customer interview synthesis:

Here are notes from [N] customer interviews. Group recurring pains,quote the exact phrases customers used, and rank themes by how manyinterviews mention them. Do not add themes that are not in the notes.---[paste notes]

Why it works: asking for exact customer phrases gives you language for copy and landing pages, and counting mentions separates patterns from one loud interview.

Customer support

24. Empathetic reply:

Draft a reply to this customer. Acknowledge the problem in onesentence, explain what we will do and when, and avoid genericapologies. Our policy: [policy]. Tone: calm, human, specific.---[paste ticket]

25. Macro from solved tickets:

From these five solved tickets about [issue], write one reusable replytemplate with placeholders for the customer's name and order details,plus a checklist the agent should verify before sending.---[paste tickets]

Why it works: building the template from real solved tickets captures the details that actually resolve the issue, and the checklist protects against sending a template blindly.

26. Help center article:

Write a help center article that answers "[question]". Steps must benumbered, one action each. Add a "Still not working?" section withthe three most common causes. Use only these product facts: [facts].

Prompts for HR, finance and operations

Back-office prompts save the most time on first drafts of documents that follow a known structure. Keep personal data out unless you are on a plan and policy that allow it.

HR and recruiting

27. Job description:

Write a job description for a [role] at a [company type, size]. Focuson outcomes in the first 6 months, not a wish list. Separate must-havefrom nice-to-have skills. Use inclusive, plain language.

28. Interview questions:

Create a structured interview for [role] covering [competencies].For each competency: two behavioral questions, what a strong answerincludes, and a red flag to listen for.

Why it works: defining what a strong answer looks like before the interview makes scoring more consistent across interviewers.

29. Onboarding plan:

Draft a 30-60-90 day onboarding plan for a new [role]. Include whothey should meet, what they should ship by day 30, and how we willknow onboarding worked.

30. Policy summary:

Summarize this policy for employees in plain language: what changes,who it affects, what they need to do and by when. Under 200 words.Flag any sentence in the original that is ambiguous.---[paste policy]

Why it works: asking ChatGPT to flag ambiguous sentences doubles as a review of the policy itself before it goes out.

Finance and operations

31. Explain a spreadsheet formula:

Explain what this formula does, step by step, then suggest a simplerversion if one exists. Assume columns are: [column list].[paste formula]

32. Variance commentary:

Here is budget vs actual for [period]. Write commentary for leadership:the three biggest variances, likely causes based on these notes, andwhat we should watch next month. Do not speculate beyond the notes.---[paste table and notes]

Why it works: the instruction not to speculate beyond your notes keeps commentary factual; you add the judgment.

33. Process documentation:

Turn this description of how we [process] into a standard operatingprocedure: purpose, roles, numbered steps, inputs and outputs, andwhat to do when something goes wrong.---[paste description]

34. Vendor comparison:

Compare these vendor quotes in a table: price, contract length,included services, exclusions and risks. Then list questions I shouldask each vendor before signing.---[paste quotes]

Prompts for product managers and engineers

Product and engineering prompts work best when you paste real artifacts: specs, logs, code and tickets. The model cannot see your systems, so give it what it needs.

Product management

35. Problem statement:

Help me write a problem statement for [feature idea]. Ask mequestions until you can state: who has the problem, how they solve ittoday, what it costs them, and how we will know we solved it.

36. User stories with acceptance criteria:

Break this feature into user stories in the format "As a [user], Iwant [action] so that [benefit]". For each, write acceptance criteriain Given/When/Then form and note edge cases.---[paste feature description]

Why it works: Given/When/Then criteria are concrete enough for engineers and testers to check, and the edge case list catches scope surprises early.

37. Release notes:

Turn this list of merged changes into customer-facing release notes.Group by benefit, not by ticket. Skip internal changes. Plain language,one sentence per item.---[paste changes]

38. Prioritization:

Score these backlog items with RICE (reach, impact, confidence,effort) using my estimates below. Show the table, then tell me whichestimates the ranking is most sensitive to.---[paste items and estimates]

Why it works: the sensitivity question shows which of your guesses matter most, so you know where better data would change the plan.

Engineering

39. Explain unfamiliar code:

Explain what this code does, line by line where it matters. Then listassumptions it makes about its inputs and anything that could fail.[paste code]

40. Debug from an error:

I get this error when [action]. Here is the stack trace and therelevant code. List the three most likely causes in order, and foreach, the quickest way to confirm or rule it out.---[paste trace and code]

Why it works: asking for ranked hypotheses with a quick test for each keeps you debugging systematically instead of applying random fixes.

41. Write tests:

Write unit tests for this function using [framework]. Cover normalcases, edge cases and invalid input. Each test must check behavior,not implementation details. Explain any case you think is a bug.[paste function]

Why it works: asking ChatGPT to test behavior and point out suspected bugs avoids tests that only restate the code.

42. Review a diff:

Review this diff as a senior engineer. List correctness issues first,then security, then readability. Be specific: file, line, problem,suggested fix. Say "no issues" for categories with none.---[paste diff]

ChatGPT is a good reviewer and explainer, but for changes across a whole repository, a coding agent that can read files and run tests does better. Our comparison of the best AI coding assistants covers those tools.

How do you save and reuse your best prompts?

Put context you repeat in custom instructions, and put context for one area of work in a ChatGPT Project. That way each new prompt can be short.

• Custom instructions apply to all your chats. On web and desktop, open Settings, then Personalization, make sure Enable customization is on, and fill in the Custom Instructions field, per OpenAI's custom instructions help page. Good content: your role, your industry, your preferred length and tone, and things to avoid.
• Projects keep related chats, files and instructions together. Select New project in the sidebar, upload reference files (style guide, product facts, templates), and add project instructions under Project settings, as described in Projects in ChatGPT.
• A personal prompt library in a notes app or doc, with your filled-in versions of the templates above, is still the fastest way to reuse a prompt across tools.

A useful custom instruction for work:

I am a [role] at a [company type]. My readers are usually [audience].Default to concise answers with the conclusion first. Use bullets forlists of three or more. Ask a clarifying question when my request isambiguous instead of guessing. Never invent numbers, names or quotes.

What should you never paste into ChatGPT at work?

Do not paste anything your company policy forbids, and treat customer data, employee data, credentials and unreleased financials as off limits unless you are sure your plan protects them.

The rules depend on your plan. According to OpenAI's help center, inputs and outputs from ChatGPT Business, Enterprise and Edu are not used for training by default. On personal Free, Plus and Pro workspaces, data sharing for training is on by default; you can switch it off under Settings, Data Controls, "Improve the model for everyone."

Even with training off, follow these habits:

• Replace names, emails and account numbers with placeholders before pasting.
• Never paste passwords, API keys or access tokens.
• Check outputs that contain facts, figures or legal language; models can state wrong things confidently. Our guide to AI hallucinations explains how to spot them.
• Keep a human decision on anything that affects a person: hiring, firing, performance or pay.

For a company-level view, including a policy template, see is it safe to use AI at work.

How to start using these prompts this week

Pick three prompts that match work you do every week and make them your own.

1. Choose one email prompt, one meeting prompt and one prompt from your role section.
2. Fill in the placeholders with a real task you have today.
3. Compare the output with what you would have written, and note what was missing from your prompt.
4. Save the improved version in your prompt library or a Project.
5. Add your role, audience and tone preferences to custom instructions.
6. After a week, keep the prompts that saved time and drop the rest.

The same habits carry over to AI agents that take on longer tasks. If your work includes code, Orkio runs coding agents such as Claude Code on an always-on server, so a well-written task keeps going after you close your laptop.

Frequently asked questions