The difference between “AI is not that useful” and “AI saves me an hour a day” is often the prompt. A vague request gets a generic answer. A clear request with context gets something you can use almost immediately.
Here are five rules for better prompts and 15 templates you can copy, adapt and reuse. They work with ChatGPT, Claude, Gemini, Microsoft Copilot and most other AI assistants.
5 rules for better business prompts
- Give context: who you are, who the audience is, what the situation is
- Be specific about the output: length, format, tone, language
- Provide the material: paste the email, notes or data instead of describing it
- Show an example of what good looks like, if you have one
- Iterate: ask for changes instead of starting over (“shorter”, “more formal”, “add a call to action”)
And one extra rule: check the result, especially facts, numbers and names.
Email and communication
1. Reply to an email
You are [role] at [company]. Write a reply to the email below.
Goal: [what you want to achieve].
Key points: [2-3 bullets].
Tone: friendly and professional. Maximum 120 words. Language: English.
Email:
[paste email]
2. Turn bullet points into a clear email
Turn these notes into a clear email to [recipient/role].
Start with the main point, then details, then the next step.
Notes:
[bullets]
3. Say no politely
Write a polite email declining [request]. Explain briefly that [reason],
offer [alternative], and keep the relationship positive. Maximum 100 words.
Meetings and documents
4. Meeting summary
Summarise the meeting notes below in this format:
1. Decisions
2. Action items (owner, deadline)
3. Open questions
Keep it under 200 words. Do not add information that is not in the notes.
Notes:
[paste transcript or notes]
See also AI meeting assistants.
5. Summarise a long document
Summarise this document for [audience] who has 2 minutes.
Give: the 5 most important points, any numbers or deadlines, and what action is needed.
Document:
[paste text]
6. Compare two documents
Compare document A and document B. List the differences in a table
with columns: topic, document A, document B, why it matters.
Sales and marketing
7. Proposal outline
Create an outline for a proposal to [client type] for [service].
Their main problem: [problem]. Our approach: [approach]. Budget range: [range].
Include: summary, problem, solution, timeline, investment, next steps.
8. Product description
Write a product description for [product] for our online store.
Target customer: [who]. Key features: [list]. Main benefit: [benefit].
Tone: [tone]. Length: 120-150 words. Add 3 bullet points with key specs.
Only use the facts provided.
9. LinkedIn post from an article
Turn the article below into a LinkedIn post for [audience].
Start with a strong first line, share 3 practical insights, end with a question.
Maximum 200 words. No hashtags in the text, add 3 relevant hashtags at the end.
Customer service
10. Customer reply using policy
You are a customer service agent for [company].
Answer the customer using only the policy below. If the policy does not cover the question, say we will check and get back within 24 hours.
Tone: empathetic, clear. Maximum 120 words.
Policy:
[paste policy]
Customer message:
[paste message]
11. Classify feedback
Classify each customer comment below into one of these categories:
[delivery, product quality, price, service, website, other].
Also mark sentiment as positive, neutral or negative.
Return a table with: comment, category, sentiment.
Analysis and planning
12. Analyse data from a table
Here is our [sales/support/marketing] data for [period].
Identify the 3 most important trends, any unusual values, and 3 questions we should investigate.
Show calculations where you use numbers.
Data:
[paste table]
Double-check any numbers. AI assistants can make calculation errors unless they use a code or analysis tool.
13. Plan a project
Create a project plan for [project] with a deadline of [date].
Team: [roles]. Break it into phases with tasks, owners and dates. List the main risks.
14. Prepare for a difficult conversation
I need to talk to [person/role] about [topic]. Their likely concerns: [concerns].
Help me prepare: key message, 3 likely objections with calm responses, and a clear next step.
15. Improve your own text
Improve the text below for clarity and readability for [audience].
Keep my meaning and tone. Use short sentences. Show the improved version, then list the main changes.
Text:
[paste text]
Build a team prompt library
Store your best prompts in a shared document or tool, with:
- The task it is for
- The prompt
- An example result
- Tips (“works best when you paste the full thread”)
This turns individual tricks into team productivity. See how to introduce AI in your team.
What not to put in prompts
Follow your company AI policy: no passwords, no sensitive personal data, no confidential information in tools that are not approved for it.
Frequently asked questions
Do prompts work the same in every AI tool?
Mostly yes. The structure (context, task, format, material) works everywhere. Results differ slightly per model.
Should prompts be in English?
Not necessarily. Modern AI assistants work well in many languages. Writing the prompt in the language of the desired output often gives the most natural result.
Is “prompt engineering” still needed?
Models are better at understanding vague requests than before, but clear context and specific output requirements still make a big difference.
Want AI built into your team’s workflows?
I help businesses move from individual prompts to integrated AI workflows connected to their own data. See my AI integration service or request a consultation.