AI Prompts for Work: 15 Copy-Paste Templates

If your AI prompts for work keep returning generic answers, the problem is rarely the model — it is the brief. This collection gives you 15 copy-paste templates for the jobs office workers actually repeat every week: emails, meetings, reports and research. Each one uses the same simple formula, so you can adapt it to ChatGPT, Claude or Gemini in under a minute.

What makes a prompt actually work at work?

Most weak prompts fail for one reason: they name a task but give the model no role, no context and no output shape. OpenAI’s official prompt-engineering guidance puts “write clear instructions” first for a reason, and Anthropic’s Claude documentation makes the same point in a different way — be explicit, add context, and show an example when the format matters. The formula below combines both sets of advice into something you can remember.

Try this: Role + Context + Task + Format + Constraints.

What happens: Instead of “Write an email about the delay”, you get a usable draft the first time.

Example — before: “Write an email about a project delay.”

Example — after: “You are a project manager writing to a client. Context: our app launch moves from 10 October to 24 October because a payment-gateway test failed. Task: write a short email that owns the delay, gives the new date, and lists the two steps we are taking to prevent a repeat. Format: subject line, 120 words maximum, calm and direct tone. Constraint: do not blame the vendor by name.”

Good to know: That second version is not “clever prompting”. It is just a proper brief — the same brief you would give a capable colleague. That is the original test behind every template below: if a smart colleague could not act on your prompt without asking three follow-up questions, the model will guess, and guessing is where generic output comes from.

Email and communication prompts (1–4)

1. The difficult email, rewritten calmly:

Rewrite the email below so it is firm but professional. Keep every fact, remove blame and sarcasm, and end with one clear next step and a date. Email: [paste email]

2. The cold email that sounds human:

You are a B2B salesperson. Write a 90-word cold email to a [role] at a [type of company]. Open with one specific observation about their business: [observation]. Offer one outcome we deliver: [outcome]. End with a low-friction question, not a meeting demand.

3. The status update from messy notes:

Turn these rough notes into a status update for [audience]. Structure: Done / In progress / Blocked / Next milestone with date. Flag anything that sounds like a risk in one sentence at the end. Notes: [paste notes]

4. The meeting follow-up:

From this meeting transcript, write a follow-up email with: decisions made, actions with an owner and a deadline for each, and open questions. If an owner or date was never stated, write ‘unassigned’ rather than inventing one. Transcript: [paste transcript]

Meetings and productivity prompts (5–8)

5. The agenda that prevents a wasted hour:

Create an agenda for a 45-minute meeting titled [title] with [attendees/roles]. Goal: [decision or outcome needed]. Give each item a time box and name the decision each item must produce. End with the pre-work each person should bring.

6. The decision brief:

Write a one-page decision brief on [decision]. Sections: Recommendation, Options considered (maximum 3), Cost and effort of each, Risks, What we will do if we are wrong. Use only the facts I provide here: [paste facts]. Mark anything you are unsure about as ‘unverified’.

7. The weekly plan from a brain dump:

Here is everything on my plate this week: [paste list]. Sort it into Must do / Should do / Delegate or drop. Schedule the Must do items into [X] focused blocks of 90 minutes. Tell me the one item you would cut first, and why.

8. The performance-review draft:

Help me draft a self-review. My role: [role]. My main results this period: [paste results with numbers]. Write 5 accomplishment bullets that start with a strong verb and keep the numbers I gave you — do not add figures I did not provide. Tone: confident, not boastful.

Research and analysis prompts (9–12)

9. The competitor teardown:

Analyse [competitor] versus [our product] for [audience]. Compare: positioning, pricing model, top 3 features, and where reviews criticise them. Put the result in a table, then give me the 2 gaps we could credibly exploit. If you do not know a fact, say so instead of estimating.

10. The document summariser with receipts:

Summarise the document below in 200 words. Then list the 5 claims in it that most need checking, quoting the exact sentence for each. Document: [paste document]

11. The spreadsheet explainer:

Here is a table of [describe data]: [paste table]. Explain in plain English: the biggest change over time, one anomaly worth investigating, and what additional data would confirm or refute your reading. Show your working in short steps.

12. The interview-prep researcher:

I am interviewing for [role] at [company]. Based on the job description below, generate the 10 most likely questions, the evidence from my CV that answers each one, and 3 smart questions I should ask them. Job description: [paste]. My CV: [paste].

Writing and content prompts (13–15)

13. The first draft with a spine:

Write a first draft of a [blog post/report/proposal] titled [title] for [audience]. Structure: hook (2 sentences), the problem, 3 supporting points with one example each, and a closing action step. Length: [word count]. Tone: [tone]. Use only these facts: [paste facts].

14. The ruthless editor:

Edit the text below. Cut 20% of the words without losing meaning, replace jargon with plain words, and list the 3 weakest sentences with a rewrite for each. Text: [paste text]

15. The repurposer:

Turn this [article/transcript] into: (a) a 5-post social thread, (b) a 100-word newsletter section, and (c) 3 headline options. Keep my original claims and wording for statistics — do not invent new numbers. Source: [paste source]

How to adapt one prompt for ChatGPT, Claude and Gemini

The same template works in all three tools; only the packaging changes slightly, based on each vendor’s own guidance.

Honest caveat: this table summarises the vendors’ published documentation as checked on 2 October 2026; it is not a benchmark test, and model behaviour changes between versions. If a prompt underperforms, add one example of a good answer before you rewrite the whole thing — both OpenAI and Anthropic rank examples among the highest-impact fixes.

Three mistakes that make every prompt worse

First, asking for everything at once. “Write my whole marketing plan” produces a shallow plan; five chained prompts — audience, offer, channels, calendar, copy — produce a usable one. Second, pasting confidential data into a consumer chatbot. Remove names, client figures and unreleased product details before you paste; use placeholders and add the real details back yourself. Third, accepting the first answer. A single follow-up — “Now make it 30% shorter and flag any claim you cannot support” — removes most of the errors that slip into AI-assisted work.

Frequently asked questions

Q: Do these prompts work in the free versions of ChatGPT, Claude and Gemini? A: Yes. All 15 use plain instructions and pasted context, so they do not need paid features. Paid tiers mainly buy you longer inputs, file uploads and newer models.

Q: How long should a work prompt be? A: Long enough to include role, context, task, format and constraints — usually 60 to 150 words. If you cannot fill in the context, the model cannot either, and it will guess.

Q: Can I save these as reusable templates? A: Yes, and you should. Save the ones you use weekly in a notes app or your tool’s saved-prompt feature, with the brackets kept as blanks to fill in.

Q: Should I tell my employer I use AI prompts at work? A: Follow your workplace policy. A safe default is to use AI for drafts and analysis, verify every fact yourself, and never paste confidential data into a personal account.

Q: Why does the same prompt give different answers in different tools? A: Each model is trained and tuned differently, which is why the adaptation table above matters — the task stays the same, but the packaging changes.

Sources and methodology

This collection was written for OpenAIMaster.ai on 2 October 2026. The prompt formula and adaptation notes rest on the vendors’ own published guidance, checked on that date: OpenAI’s prompt-engineering best practices (help.openai.com), Anthropic’s prompt-engineering overview (platform.claude.com), and the community-maintained Prompt Engineering Guide’s few-shot prompting reference (promptingguide.ai). The 15 templates themselves are original, written for this article; no performance statistics are claimed for them, and any figures quoted inside example prompts are placeholders for the reader to replace. Internal reading on this site: our coverage of the White House Accord on Super Intelligence, the report on AI agents accessing Australian government websites, and Anthropic’s reported IPO plans shows why the guardrails in these templates — verify claims, never invent owners, dates or figures — matter in real workplaces.

rangwala402

OpenAIMaster is an independent publication covering artificial intelligence — from model launches and AI news to hands-on tool reviews and practical guides. Our testing methodology is published openly, and every review is updated as tools evolve.