AI prompt templates for work: 68 copy-ready prompts for ChatGPT, Copilot, and Claude

An AI prompt template is a prompt you write once, with fixed instructions and labeled blanks for the details that change. You can’t stop ChatGPT, Copilot, or Claude from wording each answer a little differently. You can decide which parts of your request never change, like who the AI should act as, what it should know, the task, the format, and the limits.

Fix those, fill in the rest, and the output gets more predictable no matter which tool your team opens that day.

This guide is for anyone who uses ChatGPT, Copilot, or Claude at work, whether that’s in support, sales, IT, marketing, or finance, or in an industry like healthcare, legal, or insurance. It covers the five parts every template needs, 68 ChatGPT prompt templates grouped by team and industry that work in all three tools, and how to save each one as a Snippet, so the prompt that worked last week is the one you use today.

Start with the structure, because every template on this page follows it. Then copy the ones that match your team or industry.

What makes a good AI prompt template?

A good AI prompt template has five parts. It names a role, gives the context, states the task, sets the output format, and adds constraints, with [bracketed] placeholders, also called variables, for anything that changes between uses. Swap in the customer name, topic, or data, and the rest of the prompt keeps the same structure every time.

Role, context, task, output format, and constraints

• Role: Tell the AI who to act as. “Act as a customer support agent for [company name].”
• Context: Give it what it needs to know. “A customer sent this ticket: [ticket text].”
• Task: State the specific output. “Write a reply that acknowledges the issue and explains what happens next.”
• Format: Describe the shape of the answer. “Two short paragraphs. Tone: [formal / friendly / concise].”
• Constraints: Set the limits and success criteria. “Under 120 words. Don’t promise a refund. Use plain language.”

The more specific the template, the more consistent the output. A weak template like “Write a reply to this customer” lacks direction, so the AI guesses at the tone, length, and what it’s allowed to promise. A defined role, clear context slots, an output format, and a few limits give you higher-quality output you can predict, even though AI models generate slightly different wording on every run.

Prompt engineering guidance from OpenAI, Anthropic, and Microsoft

OpenAI, Anthropic, and Microsoft each publish advice for their own tools, and it lines up with the same five parts. Microsoft’s Copilot guidance names four elements of a great prompt. OpenAI’s best practices for ChatGPT and Anthropic’s prompting best practices for Claude cover the same ground in different words.

AI vendor	What its guidance asks for	Template parts it covers
Microsoft (Copilot)	Goal, context, source, and expectations	Task, context, format, and constraints
OpenAI (ChatGPT)	A persona, delimiters around pasted text, step-by-step instructions, examples, and clear, specific requests	Role, context, task, and format
Anthropic (Claude)	Clear and direct instructions, context, examples, and a role	Role, context, task, and format

Anthropic’s guide suggests you “think of Claude as a brilliant but new employee who lacks context on your norms and workflows.” That’s a useful test for any of these AI models. If a new teammate couldn’t do the task from your prompt alone, the template needs more context. For prompts where style or a consistent tone matters, include a sample output or two. Anthropic recommends 3 to 5 examples for best results.

Here’s a simple way to decide what goes in the template and what becomes a blank. Anything you’d say the same way every time belongs in the fixed layer, including the role, the format, the limits, and anything the AI should never do. Anything that changes from one use to the next becomes a blank. The fixed layer is where your best thinking lives, so it’s the part worth saving.

Before you share a new template, run it 3 to 5 times with real details and check the edge cases, like a very short input or a request the policy doesn’t cover. If a template that worked last month starts producing weaker results, tighten the format or add a context variable, and review your templates on a regular schedule, such as once a quarter. And keep one master version with one owner. Two people running slightly different versions of the same prompt can’t tell which one works better, so neither version improves.

What a filled prompt looks like

Here’s template 7 from the sales group as a full prompt, after the blanks are filled in for one prospect:

Act as an account executive. A prospect said: “Your price is higher than the tool we use now.” Write a reply under 60 words that acknowledges the concern, answers it with this fact: our annual plan includes onboarding and support at no extra cost, and ends with one question. Use only the facts given.

Only two details changed. The fixed parts carry your best thinking about how the reply should work, and the blanks carry the facts of this one conversation. That’s what makes one template useful across dozens of similar tasks.

AI prompt templates by team

Every template below is free to copy, and each one is a starting point you can adjust. Replace the bracketed variables before you send it, and keep customer names, account details, and anything confidential out of the saved version. Under each template, you’ll find a suggested Snippet abbreviation and the blanks that become Fill-ins, if you want to save it as a reusable prompt.

The 68 templates are grouped into 19 prompt categories by team, everyday task, and industry, so prompts for similar tasks sit together. Each one follows the same five parts, which makes them reusable prompt structures. Pick one template, adjust the fixed parts to your specific situation, and it becomes a reusable template for your own work.

By team: Customer support | Sales | Recruiting | Operations | Finance | IT support | Marketing | Software development

Everyday work: Writing | Research | Meetings

By industry: Healthcare | Education | Software | Legal | Retail | Hospitality | Energy and utilities | Insurance

Customer support

Support prompts work best with a firm limit on what the reply can promise, since the customer will read it as policy.

1. Ticket response

Act as a customer support agent for [company name]. A customer sent this ticket: [ticket text]. Write a reply that acknowledges the issue, explains what happens next, and gives a clear timeline. Tone: [formal / friendly / concise]. Keep it under 120 words. Don’t promise anything our policy doesn’t cover: [policy notes].

Snippet: ;aiticket. Fill-ins: company name, ticket text, tone, policy notes.

2. Escalation handoff

Act as a support team lead. Summarize this support ticket thread in 3 bullet points for a tier-2 agent: [ticket thread]. Include the customer’s core issue, what has been tried, and the customer’s current sentiment. Use only what’s in the thread.

Snippet: ;aiescalate. Fill-ins: ticket thread.

3. Proactive check-in

Write a check-in email to [customer name] about their experience with [product or resolved issue]. Keep it under 100 words. Tone: warm and professional. Don’t add offers or discounts.

Snippet: ;aicheckin. Fill-ins: customer name, product or resolved issue.

4. Knowledge base draft

Act as a technical writer. Write a knowledge base article explaining [topic] for [audience: customers / internal agents]. Format: a short intro, numbered steps, and a “What if it doesn’t work?” section at the end. Use only these steps, and flag any gap instead of guessing: [paste steps].

Snippet: ;aikb. Fill-ins: topic, audience, steps.

For 15 more, see these customer service prompts. For shared replies the whole support team can use, see TextExpander for customer support.

Sales and outreach

Sales prompts need one extra constraint. The AI should use only the facts you give it, so it doesn’t invent a customer result or a feature.

5. Cold email

Act as an account executive. Write a cold outreach email to a [prospect title] at a [industry] company. Subject line: focus on [pain point]. Body: 3 sentences max. Call to action: [specific ask]. Tone: direct, not salesy. Use only facts I’ve given you.

Snippet: ;aicold. Fill-ins: prospect title, industry, pain point, specific ask.

6. Follow-up sequence

Write a 2-email follow-up sequence for a prospect who didn’t respond to [first email topic]. Email 1 on day 3: add one useful resource or insight. Email 2 on day 7: a short, polite close. Tone: [direct / conversational]. Use only facts I’ve given you.

Snippet: ;aifollowup. Fill-ins: first email topic, tone.

7. Objection response

Act as an account executive. A prospect said: “[objection].” Write a reply under 60 words that acknowledges the concern, answers it with this fact: [fact], and ends with one question. Use only the facts given.

Snippet: ;aiobjection. Fill-ins: objection, fact.

8. Proposal summary

Summarize this proposal for a [prospect role] who has 2 minutes to read it. Highlight the business problem, the proposed solution, and the expected outcome. Use only what’s in the proposal: [proposal text].

Snippet: ;aiproposal. Fill-ins: prospect role, proposal text.

For more, see these ChatGPT prompts for sales. To see how sales teams reuse approved messaging, see TextExpander for sales.

Recruiting

Recruiting and human resources prompts carry the most risk if the AI fills gaps on its own, so these templates say what not to invent and what not to ask.

9. Job description draft

Act as a recruiter at [company name]. Draft a job description for a [job title] on the [team] team. Include a two-sentence summary, 5 responsibilities, 5 requirements split into must-have and nice-to-have, and how we work: [work location and schedule]. Use inclusive, plain language. Don’t invent benefits or salary details. Hiring manager notes: [notes].

Snippet: ;aijd. Fill-ins: company name, job title, team, work location and schedule, hiring manager notes.

10. Candidate outreach message

Write a short first message to a candidate for a [job title] role. Mention one specific thing from their background: [detail]. Explain in one sentence why the role might interest them. End with a low-pressure question. Under 80 words. Tone: warm and direct.

Snippet: ;aioutreach. Fill-ins: job title, detail.

11. Interview question set

Create 6 structured interview questions for a [job title] candidate, focused on [skills or values]. For each, add what a strong answer includes. Avoid questions about age, family, health, or other protected characteristics.

Snippet: ;aiinterview. Fill-ins: job title, skills or values.

12. Interview feedback summary

Summarize these interviewer notes into strengths, concerns, open questions, and a hire / no hire / more info recommendation with one sentence of reasoning. Use only what’s in the notes. Notes: [paste notes, with the candidate’s name removed].

Snippet: ;aifeedback. Fill-ins: notes.

For finished messages instead of prompts, see these text templates for recruiters.

Operations and admin

Operations prompts turn rough ideas and notes into documents other people rely on, so each one asks the AI to flag what’s unclear.

13. Process documentation (SOP) draft

Act as an operations lead. Turn these rough notes into a step-by-step procedure for [process name]. Format: purpose (one sentence), who does it, numbered steps, and “If something goes wrong”. Flag any step that looks unclear instead of guessing. Notes: [paste notes].

Snippet: ;aisop. Fill-ins: process name, notes.

14. Vendor comparison

Compare [vendor A] and [vendor B] for [use case] using only the information below. Format: a table with price, contract terms, support, and risks, then a two-sentence recommendation that names the main trade-off. Information: [paste details].

Snippet: ;aivendor. Fill-ins: vendor A, vendor B, use case, details.

15. Policy change announcement

Write an internal announcement about this change: [what’s changing]. Include what changes, when it starts ([date]), who it affects, and what people need to do. Under 150 words. Tone: clear and calm.

Snippet: ;aipolicy. Fill-ins: what’s changing, date.

16. Incident summary

Summarize this incident for [audience: leadership / the team / customers]: [paste timeline]. Format: what happened, impact, what we did, and what we’ll change. Keep it factual and blame-free.

Snippet: ;aiincident. Fill-ins: audience, timeline.

To see where reusable text fits in other admin work, see TextExpander for operations teams.

Finance

Finance prompts deal with real money and real customers, so each one tells the AI to use only the numbers and policy you paste in.

17. Overdue invoice reminder

Act as an accounts receivable specialist at [company name]. Write a polite payment reminder to [customer name] for invoice [invoice number] for [amount], due [due date]. Explain how to pay: [payment options]. Tone: [friendly / firm]. Keep it under 120 words. Don’t mention late fees unless they’re in these terms: [terms].

Snippet: ;aiinvoice. Fill-ins: company name, customer name, invoice number, amount, due date, payment options, tone, terms.

18. Expense policy answer

Act as a member of the finance team. An employee asked: “[question].” Answer using only this section of our expense policy: [paste policy]. If the policy doesn’t cover it, say so and tell them who to ask: [contact]. Under 100 words.

Snippet: ;aiexpense. Fill-ins: question, policy section, contact.

19. Budget variance explanation

Explain this budget variance for [audience: department heads / leadership]: [paste budget vs. actual figures]. Format: the 3 biggest variances, a likely cause for each based only on these notes: [notes], and one question to confirm with the budget owner. Don’t invent causes.

Snippet: ;aivariance. Fill-ins: audience, figures, notes.

20. Month-end close summary

Summarize this month-end close status for the finance team: [paste checklist status]. Format: completed, in progress, blocked (with owner), and what’s due next week. Under 150 words.

Snippet: ;aiclose. Fill-ins: checklist status.

To see how finance teams reuse approved wording, see TextExpander for finance.

IT support

IT support prompts work best when they stop short of the fix. The AI drafts the reply or sorts the ticket, and a technician makes the call.

21. Access request reply

Act as an IT support technician. Reply to this access request: [request]. Explain what we need to approve it: [approval steps], and how long it usually takes: [timeframe]. Never ask the user to send a password. Under 100 words.

Snippet: ;aiaccess. Fill-ins: request, approval steps, timeframe.

22. Outage status update

Write an internal status update about [system] being unavailable. Include what’s affected, when it started ([time]), the workaround if there is one: [workaround], and when the next update will come: [next update time]. Keep it calm and under 120 words. Don’t guess at the cause.

Snippet: ;aioutage. Fill-ins: system, time, workaround, next update time.

23. End-user troubleshooting steps

Turn these technician notes into numbered troubleshooting steps a non-technical employee can follow for [problem]: [paste notes]. Use plain language and one action per step. End with when to contact IT: [contact].

Snippet: ;aitroubleshoot. Fill-ins: problem, notes, contact.

24. Ticket triage

Read this IT ticket: [ticket text]. Suggest a category from this list: [categories], a priority (low, medium, or high) with one sentence of reasoning, and the first question to ask the user. Don’t attempt the fix.

Snippet: ;aitriage. Fill-ins: ticket text, categories.

For more, see TextExpander for IT support.

Marketing

Marketing prompts include a newsletter writer, a LinkedIn post generator, and a prompt for image generation tools. Each one limits the AI to the claims you’ve approved.

25. Newsletter writer

Act as a newsletter writer for [company name]. Draft this month’s newsletter for [audience] using these updates: [paste updates]. Format: a subject line, a two-sentence intro, 3 short sections with headings, and one call to action: [CTA]. Tone: [tone]. Under 400 words. Use only the facts given.

Snippet: ;ainewsletter. Fill-ins: company name, audience, updates, CTA, tone.

26. LinkedIn post generator

Act as a LinkedIn ghostwriter for [name and role]. Write 3 versions of a LinkedIn post about [topic] for [target audience]. Each version: a hook in the first line, under 150 words, one specific example from this source material: [paste notes], and a closing question. No hashtags or buzzwords.

Snippet: ;ailinkedin. Fill-ins: name and role, topic, target audience, source material.

27. Image generation prompt

Write a prompt for an AI image generation tool to create a [blog header / social graphic / product mockup] about [subject]. Include the subject, setting, style: [style], color palette: [brand colors], composition, and aspect ratio: [ratio]. List what to avoid: text in the image, logos, and distorted hands.

Snippet: ;aiimage. Fill-ins: image type, subject, style, brand colors, ratio.

28. Ad copy variations

Write 5 ad copy variations for [product or offer] on [platform] for [target audience]. Each: a headline under [character limit] characters and a description under [character limit] characters. Use only these approved claims: [claims]. Vary the angle across problem, outcome, proof, curiosity, and timing.

Snippet: ;aiadcopy. Fill-ins: product or offer, platform, target audience, character limits, claims.

For more, see TextExpander for marketing.

Software development and code review

Developers use structured prompts for code review, debugging, and documentation, where a clear output format matters as much as it does in a support reply.

29. Code review

Act as a senior software engineer. Review this code for bugs, security issues, and readability problems: [paste code]. For each issue, give the line number, what’s wrong, and a suggested fix. If you aren’t sure something is a bug, say so. Don’t rewrite code that already works.

Snippet: ;aicodereview. Fill-ins: code.

30. Debugging help

Act as a senior developer. I’m getting this error: [error message]. Here’s the relevant code: [paste code], and here’s what I’ve tried: [attempts]. Explain the most likely cause, then give a suggested fix. If you need more information, ask for it instead of guessing.

Snippet: ;aidebug. Fill-ins: error message, code, attempts.

31. Structured data as valid JSON

Extract these fields from the text below and return only valid JSON that matches this schema: [JSON schema]. Use null for any field that isn’t in the text. Don’t add fields, comments, or explanations. Text: [paste text].

Snippet: ;aijson. Fill-ins: JSON schema, text.

32. Pull request description

Write a pull request description from this summary of changes: [paste summary]. Format: what changed, why, how to test it, and any risks. Under 200 words. Use plain language a reviewer from another team can follow.

Snippet: ;aipr. Fill-ins: summary of changes.

For more, see TextExpander for software development.

Everyday prompt categories: writing, research, and meetings

Some prompts don’t belong to one team. These cover the writing, research, and meeting work almost everyone does, whatever their role.

Writing and content creation

Writing prompts give the AI the audience and length up front, which matters more for content creation than clever wording. The same pattern works for a blog post outline or the copy for landing pages. Give the AI the keyword, the target audience, and the search intent, then ask for headings with one line each.

33. Email draft

Write a [type: cold outreach / follow-up / announcement] email to [audience] about [topic]. Tone: [professional / casual / concise]. Length: [word count]. Points to include: [key points].

Snippet: ;aiemail. Fill-ins: type, audience, topic, tone, word count, key points.

34. Rewrite for a different audience

Rewrite the text below for [audience]. Keep every fact and number the same. Adjust the vocabulary, length, and tone to fit: [tone]. Target length: [word count]. Text: [paste text].

Snippet: ;airewrite. Fill-ins: audience, tone, word count, text.

35. Editing pass

Edit the following text for clarity and conciseness. Remove filler words. Preserve the original meaning. Target length: [word count]. Text: [paste text].

Snippet: ;aiedit. Fill-ins: word count, text.

36. Summarize a long document

Summarize the document below for [audience] who has 2 minutes to read it. Format: a one-sentence bottom line, 3 to 5 key points, and any decisions or deadlines. Quote numbers exactly. Document: [paste text].

Snippet: ;aisummary. Fill-ins: audience, document.

Template 33 is a starting point for email, and the LinkedIn post generator in the marketing group covers social posts. For 12 email-specific versions, see these ChatGPT prompts for email.

Research and analysis

Research prompts are where AI is most likely to sound sure about something it doesn’t know. Ask for sources or limits, remove confidential data before you paste anything, and follow your company’s AI use policy.

37. Competitive summary

Summarize what makes [competitor name] appealing to [target audience]. Cover positioning, key differentiators, and commonly cited weaknesses. Use only publicly available information, and say where you aren’t sure.

Snippet: ;aicompete. Fill-ins: competitor name, target audience.

38. Pros and cons

List the pros and cons of [decision / tool / approach] for a [role] at a [company size] company. Format: two columns, 5 items each. Then name the biggest trade-off in one sentence.

Snippet: ;aiproscons. Fill-ins: decision, role, company size.

39. Audience research

Act as a market researcher. Describe the top 5 concerns of [buyer persona] when evaluating [product type or solution]. Include the specific language they would use.

Snippet: ;aiaudience. Fill-ins: buyer persona, product type.

40. Data interpretation

Interpret the following data and summarize the key takeaway in 2 to 3 sentences for a non-technical audience. If the data doesn’t support a conclusion, say so. Data: [paste data].

Snippet: ;aidata. Fill-ins: data.

Meetings and summaries

Meeting prompts work well as templates because the source material changes every week and the format never does.

41. Agenda builder

Create a meeting agenda for a [meeting type] with [attendees or roles]. Goal: [goal]. Time available: [duration]. Include a time allocation for each agenda item.

Snippet: ;aiagenda. Fill-ins: meeting type, attendees, goal, duration.

42. Decision brief

Write a one-page decision brief on [topic]. Structure: Background, Options Considered, Recommendation, Risks, and Next Steps. Audience: [executive / team / stakeholder].

Snippet: ;aidecision. Fill-ins: topic, audience.

43. Weekly status update

Write a weekly status update for [project name]. Format: three sections (What was completed, What’s in progress, and Blockers). Use these notes: [paste notes].

Snippet: ;aistatus. Fill-ins: project name, notes.

44. Action item extraction

Extract every action item from these meeting notes. For each, list the owner, the task, and the due date if one is mentioned. If an owner is unclear, say so. Notes: [paste notes].

Snippet: ;aiactions. Fill-ins: meeting notes.

AI prompt templates by industry

Industry templates use the same five parts, with stricter constraints in the fixed layer. Treat them as starting points to customize with your own requirements, and check what applies to your organization before you use them. A few rules are worth building into every template for these industries:

• Healthcare: keep protected health information out of saved prompts, and out of any AI tool your organization hasn’t approved for it.
• Legal: ask for explanations and first drafts, not legal advice, and remove client names before you paste anything.
• Insurance and energy: tell the AI not to make coverage, credit, or payment decisions.
• Retail and hospitality: paste your actual policy, so replies don’t invent refunds or compensation.

Healthcare

These templates keep patient details in Fill-ins and out of the saved prompt.

45. Appointment reminder

Write a friendly appointment reminder for a [practice type] patient. Include the date and time: [appointment date and time], location: [location], what to bring: [items], and how to reschedule: [phone or link]. Under 80 words. Don’t include any medical details.

Snippet: ;aiapptremind. Fill-ins: practice type, appointment date and time, location, items, phone or link.

46. Plain-language patient instructions

Rewrite these care instructions at a 6th-grade reading level: [paste instructions]. Keep every step, dose, and time exactly as written. Use short numbered steps. End with a line telling the patient to call [contact] with questions.

Snippet: ;aipatientinfo. Fill-ins: instructions, contact.

47. Prior authorization summary

Using these de-identified notes, draft a summary for a prior authorization request: [paste notes]. Format: requested service, diagnosis code, relevant history, and why it’s medically necessary, based only on the notes. Flag anything payers usually ask for that’s missing.

Snippet: ;aipriorauth. Fill-ins: de-identified notes.

For more, see TextExpander for healthcare.

Education

Education prompts save the most time on the documents teachers write every term.

48. Lesson plan outline

Act as a [grade level] [subject] teacher. Create a lesson plan on [topic] that fits a [class length] class. Include a learning objective, a warm-up, the main activity, a check for understanding, and homework. Align it to this standard: [standard].

Snippet: ;ailesson. Fill-ins: grade level, subject, topic, class length, standard.

49. Family update

Write a short update to the families of [class or grade] about [topic]. Include what’s happening, the date: [date], and what families need to do: [action]. Under 120 words. Warm, clear tone, no jargon.

Snippet: ;aifamily. Fill-ins: class or grade, topic, date, action.

50. Grading rubric

Create a 4-level grading rubric for [assignment] in [subject] for [grade level]. Criteria: [criteria]. Describe each level in one sentence per criterion, in language students can understand.

Snippet: ;airubric. Fill-ins: assignment, subject, grade level, criteria.

For more, see TextExpander for education.

Software

Software companies send a steady stream of release notes, announcements, and check-ins, which makes them good candidates for templates.

51. Release notes

Turn this list of changes into customer-facing release notes for [product name] version [version]: [paste changes]. Group them under New, Improved, and Fixed. Write one plain-language sentence per item about what users can now do. Leave out internal ticket numbers.

Snippet: ;aireleasenotes. Fill-ins: product name, version, changes.

52. Feature announcement email

Write an email announcing [feature] to [customer segment]. Explain the problem it solves, how to use it in 3 steps: [steps], and where to learn more: [link]. Under 150 words. Use only these facts: [facts].

Snippet: ;aifeature. Fill-ins: feature, customer segment, steps, link, facts.

53. Customer success check-in

Act as a customer success manager at a software company. Write a check-in to [customer name], whose product usage has dropped. Context: [account notes]. Ask one open question about their goals and offer this help: [specific help]. Under 100 words. Don’t sound like an automated alert.

Snippet: ;aicsm. Fill-ins: customer name, account notes, specific help.

For more, see TextExpander for software companies.

Legal

Legal prompts help with first drafts and plain-language explanations. Remove client names and confidential details first, and follow your firm’s AI policy.

54. Plain-language clause summary

Explain this contract clause in plain language for a non-lawyer: [paste clause, with names removed]. Format: what it says, what it requires of each party, and questions to raise with an attorney. Don’t give legal advice or say whether the clause is enforceable.

Snippet: ;aiclause. Fill-ins: clause.

55. Client intake follow-up

Write a follow-up email to a prospective client after an intake call about [matter type]. List the documents we need: [documents], the next step: [next step], and the deadline: [date]. Under 150 words. Professional and reassuring. Don’t make promises about outcomes.

Snippet: ;aiintake. Fill-ins: matter type, documents, next step, date.

56. Matter status update

Summarize the status of this matter for the client in plain language: [paste internal notes]. Format: what happened since the last update, what’s next, and anything we need from them. Under 150 words. Leave out internal strategy.

Snippet: ;aimatter. Fill-ins: internal notes.

For more, see TextExpander for legal teams.

Retail

Retail prompts cover the product pages and customer questions that come up every day.

57. Product description

Write a product description for [product name] for our online store. Audience: [target customer]. Include 3 benefits based on these features: [features], materials and sizing: [details], and care instructions: [care]. Under 120 words. Don’t invent specifications.

Snippet: ;aiproduct. Fill-ins: product name, target customer, features, details, care.

58. Out-of-stock reply

Reply to a customer asking about [product], which is out of stock. Give the restock date if we know it: [date or unknown], suggest these alternatives: [alternatives], and offer to let them know when it’s back. Under 80 words. Friendly tone.

Snippet: ;aioutofstock. Fill-ins: product, date, alternatives.

59. Return request reply

Reply to this return request: [customer message]. Apply only this return policy: [policy]. If the return qualifies, list the next steps: [steps]. If it doesn’t, explain why kindly and offer this option: [alternative]. Under 120 words.

Snippet: ;aireturn. Fill-ins: customer message, policy, steps, alternative.

For more, see TextExpander for retail.

Hospitality

Hospitality prompts keep guest messages warm and specific without promising what the property can’t deliver.

60. Pre-arrival email

Write a pre-arrival email to [guest name], staying at [property] from [check-in date] to [check-out date]. Include the check-in time: [time], parking: [parking details], and one local recommendation: [recommendation]. Under 120 words. Warm, welcoming tone.

Snippet: ;aiprearrival. Fill-ins: guest name, property, dates, time, parking details, recommendation.

61. Review response

Write a public reply to this guest review: [review]. Thank the guest, acknowledge the specific issue without making excuses, say what we’re doing about it: [action], and invite them to contact [contact]. Under 100 words. Don’t offer compensation in a public reply.

Snippet: ;aireview. Fill-ins: review, action, contact.

62. Group booking reply

Reply to this group booking inquiry: [inquiry]. Confirm availability for [dates] based on: [availability], and share the group rate: [rate], what’s included: [inclusions], and the deposit deadline: [deadline]. Under 150 words.

Snippet: ;aigroup. Fill-ins: inquiry, dates, availability, rate, inclusions, deadline.

For more, see TextExpander for hospitality.

Energy and utilities

Utility messages reach thousands of customers at once, so these templates ask for calm, plain language and no promised credits.

63. Planned outage notice

Write a customer notice about a planned [service] outage in [area] on [date] from [start time] to [end time]. Explain why: [reason], how to prepare, and where to get updates: [link or phone]. Under 120 words. Calm, plain language.

Snippet: ;aiplannedoutage. Fill-ins: service, area, date, start and end time, reason, link or phone.

64. High bill inquiry reply

Reply to a customer asking why their bill is higher than usual: [customer message]. Explain the likely reasons based only on this account summary: [usage and rate notes], and offer these options: [budget billing / payment plan / energy audit]. Under 150 words. Don’t promise a credit.

Snippet: ;aihighbill. Fill-ins: customer message, usage and rate notes, options.

65. Payment arrangement confirmation

Write a confirmation of a payment arrangement for the account ending in [last four digits]: [payment schedule]. Explain what happens if a payment is missed, per this policy: [policy], and how to contact us. Under 120 words. Respectful tone.

Snippet: ;aipaymentplan. Fill-ins: last four digits, payment schedule, policy.

For more, see TextExpander for energy and utilities.

Insurance

Insurance prompts explain and update. They never decide what’s covered.

66. Claim status update

Write a claim status update to [policyholder name] for claim [claim number]. Current status: [status]. Next step: [next step], expected timing: [timing], and documents still needed: [documents]. Under 120 words. Don’t say whether the claim will be paid.

Snippet: ;aiclaim. Fill-ins: policyholder name, claim number, status, next step, timing, documents.

67. Renewal reminder

Write a policy renewal reminder for a [policy type] policy expiring on [date]. Mention what changes at renewal: [changes], how to renew: [steps], and who to contact: [agent name]. Under 120 words. Use only the facts given.

Snippet: ;airenewal. Fill-ins: policy type, date, changes, steps, agent name.

68. Coverage term explainer

Explain the insurance term [term] in plain language for a policyholder, with one everyday example. Under 100 words. End with a line telling them to check their policy documents or ask [agent name] how it applies to them. Don’t make coverage decisions.

Snippet: ;aicoverage. Fill-ins: term, agent name.

For more, see TextExpander for insurance.

How to use these ChatGPT prompt templates in Copilot and Claude

Every template on this page works in ChatGPT, Microsoft Copilot, and Claude, because all three respond to the same five parts. The wording of each answer will differ a little between tools, and between runs in the same tool. The structure carries over, so a filled prompt that works in one tool is a good starting point in the others.

Where context lives in your AI workflows

What differs is where each tool keeps background context. ChatGPT keeps details about you and your work in custom instructions and Projects. Claude has Projects, where you can add files and instructions that apply to every chat. Copilot can draw on Microsoft 365 files you point it to as source material.

In most AI workflows, use those features for long-lived context, like who you are, what your company sells, and the style guide you follow. Keep the reusable prompt templates themselves somewhere you control. AI platforms rename and retire features on their own schedule. OpenAI, for example, is retiring custom GPTs on December 11, 2026. A prompt saved as a Snippet works in whichever AI tool or app you type in, so a change like that doesn’t take your best prompts with it.

One tip from Anthropic’s guidance works in all three tools. When you paste a long document, put it above your question, not below it.

If you mostly work in ChatGPT, see how to save prompts in ChatGPT for the built-in options and where each one stops.

Turn a template into a Snippet with Fill-ins

A template only helps if it’s there when you need it. Saving each AI prompt template as a TextExpander Snippet puts it a short abbreviation away in ChatGPT, Copilot, Claude, or any other app where you type, so you don’t hunt through a doc or write a new prompt from memory.

TextExpander solves the frustrating problem of repetitive AI prompt input. Instead of constantly retyping or copying and pasting the same prompts throughout the day, TextExpander makes them instantly accessible through shortcuts.

Koji Noguchi, Attorney at Law, OFS Fitel, LLC

Save a new prompt as a Snippet in four steps

Here’s how to set one up:

1. Copy a template from this page into a new Snippet.
2. Give it an abbreviation that starts with a semicolon and a shared prefix, such as ;aiticket or ;aijd, so your prompts are easy to find and don’t expand by accident.
3. Replace each bracketed variable with a Fill-in field, so TextExpander asks for that detail every time the Snippet expands.
4. Type the abbreviation in ChatGPT, Copilot, or Claude, complete the fields, and send the full prompt.

For example, ;aiactions holds template 44. When it expands, one Fill-in field asks for the meeting notes, and the rest of the prompt arrives exactly as you tested it. The ;aiticket Snippet asks for four details (the company name, the ticket, the tone, and your policy notes) before it fills in the formatted prompt.

Keep sensitive data out of the full prompt

Keep customer names, account details, and confidential data out of the saved template. Enter them in a Fill-in at the moment of use, and only where your company’s AI use policy allows.

If your team handles regulated or confidential information, check how every SaaS tool in the workflow handles security before anyone saves prompts in it, including the AI tools themselves. TextExpander maintains SOC 2 and SOC 3 certification, is compliant with HIPAA Security, Privacy, and Breach Notification rules, with BAAs available on qualified Enterprise plans, and never uses customer data to train AI models. Snippets still shouldn’t contain patient or customer personal data, which is why those details belong in Fill-ins.

Plenty of TextExpander customers already work this way. Mario Jimenez, a Merchant Experience Partner at DoorDash, uses TextExpander for “fast emails, quick AI prompts, and case notes.” To see how the same approach scales to a larger prompt library, see the TextExpander AI prompt manager.

Share one approved version with your team

A shared doc or a pinned message works for a few weeks. Then someone pastes an older copy, someone else edits their own version, and the team ends up running three slightly different versions of the same cold email prompt. Nobody knows which one works best, and the newest hire is using last quarter’s.

Reply templates drift the same way. Amwell’s support team moved from individually maintained AutoHotKey templates to shared Snippets after seeing the risk. As their Director of Support & Operations put it, “there was a chance of someone accidentally using an old version and sending out bad information.”

Keep AI outputs consistent with one template library

A shared Snippet Group works the same way for prompts. One owner maintains each group, and when they improve a prompt, the update reaches everyone who uses it. On Business and Enterprise plans, teammates can suggest a change through Snippet Requests, and the owner approves it before it goes live. That keeps one prompt library for the team instead of a personal copy per person, and it cuts the copy-and-paste errors that creep in when everyone keeps their own version. Output consistency improves for a simple reason. Everyone starts from the same words.

One prompt library for people and AI workflows

The same library can serve AI tools too. The TextExpander MCP Server, in early access, lets Claude, ChatGPT, and Microsoft Copilot read from the approved Snippet library your team uses. AI creates. TextExpander deploys. In practice, AI is great at generating the first version of a prompt, and TextExpander keeps the version your team approved available wherever anyone types.

For a step-by-step plan, see how to build an AI prompt library your team will keep using. To compare options, see these AI prompt management tools, or see how shared Snippet libraries for teams work.

Build your prompt templates once, then reuse them

The fastest way to start is with the three prompt examples that match your team or industry. Run each one a few times with real details, tighten the format and limits until the output is something you’d send, then save it as a Snippet. That tested version is the one worth keeping.

AI tools will keep getting better at drafting, and they’ll keep renaming and retiring features along the way. The decision about what a good prompt says only has to be made once. Fix the parts that shouldn’t change, leave blanks for the rest, and let AI handle what’s new each time. When your team is ready, share the group so everyone starts from the same high-quality prompts. Treat them as reusable assets, the same way you’d treat an approved email template. Start with a free TextExpander account.