Prompt Templates for Everyday Business Tasks

Prompt templates for business are reusable instructions with placeholders that you fill in each time, such as [paste email] or [audience]. They save time and, more importantly, make results consistent across a team. Below are twelve templates for common tasks: emails, summaries, meeting notes, extraction, reports, feedback and planning. Each follows the same structure of task, context, audience, format and rules described in our prompt engineering guide. Copy them, replace the brackets, and adapt the rules to your company.

If you prefer filling in fields over editing text, the AI prompt generator builds prompts with the same structure and includes several of these templates as presets.

How to use these templates

• Replace everything in square brackets.
• Keep the triple quotes around pasted material; they separate your content from the instructions.
• Delete rules that do not apply and add your own.
• Save the versions that work well in a shared place so the team uses the same wording.
• Follow your company's AI policy on what data may be pasted into which tool.

1. Reply to a customer email

You are a customer service representative for [company], which [one sentence about what you do].

Task: Write a reply to the customer email below. Acknowledge the issue, explain the
next step and give a realistic timeframe.

Facts you may use:
"""
[order details, relevant policy excerpt]
"""

Customer email:
"""
[paste email]
"""

Rules:
- Use only the facts above. If something is missing, write [CHECK] instead of guessing.
- Do not promise refunds, discounts or dates that are not in the facts.
- [Formal/informal] address, friendly and clear, under 150 words.
- End with one clear next step for the customer.

2. Summarise a document for a decision-maker

Summarise the document below for [role, e.g. the managing director], who needs to decide
[what decision].

Structure:
1. One-sentence summary
2. Key points (max. 5 bullets)
3. Risks or open questions
4. What a decision requires

Rules: Use only information from the document. Quote figures exactly. Mark anything
unclear as an open question. Maximum 250 words.

"""
[paste document]
"""

3. Meeting notes to action items

Turn the meeting notes below into a summary for people who did not attend.

Output:
- Decisions made
- Action items as a table: action | owner | due date
- Open questions

If an owner or due date is not stated, write "not specified". Do not add actions that
were not discussed.

"""
[paste notes or transcript]
"""

4. Extract structured data from a document

Extract the following fields from the document below and return valid JSON only:
{"supplier_name": "", "invoice_number": "", "invoice_date": "YYYY-MM-DD",
 "net_amount": 0, "tax_amount": 0, "currency": ""}

Rules:
- Use null for any field that is not present. Do not estimate.
- Amounts as numbers without currency symbols.
- No text outside the JSON.

"""
[paste document text]
"""

For many documents in one batch, process them one per request; results are easier to check.

5. Rewrite for clarity

Rewrite the text below so it is easier to read for [audience].

Keep: the meaning, all facts, names, numbers and dates.
Change: long sentences, jargon, passive constructions.
Format: [same structure / short paragraphs / bullet points].
After the text, list the three most important changes you made.

"""
[paste text]
"""

6. Adapt content for another channel

Adapt the content below for [channel, e.g. a LinkedIn post / a customer newsletter /
an internal announcement].

Audience: [who]
Length: [limit]
Tone: [e.g. professional and approachable]
Must include: [key message, link placeholder, call to action]
Avoid: exaggerated claims, emojis [if applicable], hashtags [if applicable].

"""
[paste original content]
"""

7. First draft of a report section

You are an analyst writing for [audience].

Write the section "[section title]" of a [type of report] based only on the data and
notes below.

Structure: short introduction, findings with figures, implications, next steps.
Length: about [number] words.
Rules: Separate observations from interpretation. Do not invent numbers. Flag
assumptions explicitly.

Data and notes:
"""
[paste]
"""

8. Analyse customer feedback

Analyse the customer feedback below.

Output:
1. A table of themes: theme | short description | number of mentions | example quote
2. The three themes that most need action, with reasons
3. Any feedback that needs an individual response (e.g. complaints, safety issues)

Rules: Count only what is in the data. Use exact quotes. Do not identify individuals
in the summary.

"""
[paste feedback, one item per line]
"""

9. Prepare for a difficult conversation

I need to have a conversation with [role, e.g. a supplier] about [topic].
Context: [what happened, what I want to achieve, constraints].

Help me prepare:
- My key message in two sentences
- Three likely objections and calm, factual responses
- Questions I should ask
- What a good outcome and an acceptable compromise look like

Keep it practical and concise.

10. Write a job description draft

Write a job description for the role [title] at [company type and size].

Include: purpose of the role, main responsibilities (max. 7), required skills,
nice-to-have skills, what we offer [facts], how to apply.

Rules: Inclusive, plain language. No exaggerated claims. Do not include requirements
not listed below.

Notes from the hiring manager:
"""
[paste notes]
"""

Note that using AI to draft a job ad is very different from using it to evaluate applicants. The second can be a high-risk use under the EU AI Act; see EU AI Act risk categories.

11. Brainstorm with constraints

Suggest [number] ideas for [goal].

Context: [business, audience, budget, what has been tried]

For each idea give: name, one-sentence description, main benefit, main risk,
rough effort (low/medium/high).

Rules: Ideas must be realistic for [team size / budget]. Avoid generic advice.
Rank from most to least promising and explain the top choice in two sentences.

12. Review a draft against a checklist

Review the draft below against these criteria:
[list 3–7 criteria, e.g. "every claim is supported by the facts", "no promises about
delivery dates", "under 200 words"]

For each criterion: pass or fail, with the exact passage if it fails, and a suggested fix.
Do not rewrite the whole draft.

Facts:
"""
[paste]
"""

Draft:
"""
[paste]
"""

This template is useful as a second step after any drafting template, and as a support for human review. It does not replace the reviewer; see human-in-the-loop AI for how to design that step.

Making templates work for your team

Add this	Why
Company and product facts	So the AI does not fill gaps with assumptions
Form of address and tone rules	Consistency with your brand
A list of things never to promise	Reduces risky output
One example of excellent output	Often more effective than describing style; see few-shot prompting
Owner and version	So improvements are shared and tracked

When the same template is used in an automated workflow rather than a chat, its standing instructions belong in a system prompt. How to write a system prompt explains the structure.

Common mistakes with templates

• Leaving placeholders unfilled. The model may treat "[audience]" literally or invent something.
• Pasting too much. Long irrelevant material dilutes focus. Include what the task needs.
• Copying templates without adapting rules. Generic rules give generic results.
• Never testing. Try each template on several real cases, including awkward ones, before sharing it.
• Trusting the output unchecked. Templates improve consistency, not correctness.

Next step

Pick the two templates closest to tasks you do every week, adapt them, and test each on five real cases. Or build your own from scratch with the AI prompt generator, which assembles role, task, context, format and rules into a ready-to-copy prompt.