Seven prompts covering the situations that come up most in professional work. Each is built on the CLEAR framework, so you can see the structure and adapt it rather than copying blindly.
16 min read
7 prompts
Copy and adapt
A prompt library is not a set of magic words. It is a set of well-structured starting points that save you from rebuilding the same scaffolding every time.
How to use these
Replace anything in [square brackets] with your specifics. The more precisely you fill those in, the better the result — a prompt with lazy placeholders produces lazy output, regardless of how good the structure is.
Before you paste anything
Remove client names, colleague names, personal data, contract terms, and anything company-confidential. Substitute generic descriptors — "a major client" rather than the actual name. Check your organisation's AI policy before using these on work material.
1. The learning prompt
Use when: you need to understand something new quickly and properly, rather than collecting a vague impression of it.
Learning
Teach me [topic] as if I am a complete beginner.
Explain it in simple language, provide one real-world
example, mention the common mistakes people make, and
give me one small practical exercise I can complete in
under 15 minutes.
After the explanation, ask me two questions to check
whether I actually understood it.
The final line is what makes this work. Being questioned exposes the gap between recognising an explanation and understanding it — a gap that passive reading never reveals.
2. The career prompt
Use when: you want role-specific direction rather than generic advice about "learning AI".
Career
Act as an AI career coach.
My current role is [role].
I have [number] years of experience.
My key responsibilities are [responsibilities].
My main career goal is [goal].
Identify the five most relevant AI skills for my role.
For each skill, explain:
1. Why it matters for my specific role
2. How I would use it in my actual work
3. Which tool I should practise it with
4. One beginner exercise to build it
5. How I would measure improvement
Rank them by which would create visible value soonest.
Then tell me which two I should ignore for now, and why.
The instruction to name what you should ignore is deliberately included. Most advice only adds; a genuinely useful roadmap also subtracts.
3. The productivity prompt
Use when: you have a recurring task and suspect it could be faster or better.
Productivity
Review the following task:
[Describe the task, how often you do it, and how
long it currently takes.]
Suggest how AI could help me complete it faster or
improve the quality.
Provide:
1. The recommended approach
2. A step-by-step workflow
3. A ready-to-use prompt I can copy
4. The risks or mistakes to watch for
5. A method to measure the time actually saved
Be honest if AI is a poor fit for this task and say so.
Why that last line matters
Without explicit permission to say "this is not a good fit", models will find a way to be helpful about tasks AI genuinely should not touch. Inviting the negative answer makes the positive answers more trustworthy.
4. The email prompt
Use when: a message is sensitive, political, or you have rewritten it four times already.
Email
Rewrite the following email so it sounds clear,
professional, respectful, and concise.
Recipient: [Their role and how they prefer to be
communicated with.]
Goal: [What you want to happen after they read it.]
Constraints: [Anything you must or must not say.]
Original email:
[Paste your draft.]
Keep my voice — do not make it sound corporate.
Then list anything in my original that could be
misread, and explain why.
Ask for the misreading analysis even when you do not use the rewrite. It frequently catches a sentence that lands differently than you intended.
5. The meeting prompt
Use when: you have messy notes and need structured output others can act on.
Meetings
Convert the following meeting notes into:
- A three-sentence summary
- Key decisions made
- Action items with owners and deadlines
- Risks or blockers raised
- Open questions that were not resolved
- Follow-up questions I should ask
If an action item has no clear owner or deadline in the
notes, list it separately under "Needs clarification"
rather than guessing.
Meeting notes:
[Paste notes.]
The "Needs clarification" instruction prevents the most damaging failure mode here — inventing an owner for a task nobody actually agreed to take.
6. The research prompt
Use when: you are exploring an unfamiliar area and need to know what is solid versus what is assumed.
Research
Help me research [topic].
Separate your response clearly into:
- Verified information you are confident about
- Assumptions or generalisations you are making
- Areas that genuinely need further validation
- Practical implications for [my situation]
- Recommended next actions
For anything in the first section, tell me what kind of
source would confirm it. Do not present uncertain claims
as established fact.
This reduces risk but does not remove it
A model's own confidence estimate is itself unreliable. This structure helps you triage what to check first — it is not a substitute for checking. Verify anything you will act on, cite, or send.
7. The problem-solving prompt
Use when: you are stuck and suspect you are looking at the problem wrongly.
Problem-solving
Act as a structured problem-solving consultant.
Problem:
[Describe the problem, what you have already
tried, and what constraints you are working within.]
Help me identify:
1. Possible root causes, ordered by likelihood
2. Information I am missing that would change the answer
3. Potential solutions with the trade-offs of each
4. The risks of each solution
5. The single best next action
Before answering, tell me if I have described the problem
in a way that assumes a particular cause.
That closing instruction regularly produces the most valuable output — being told your framing is loaded is more useful than a well-reasoned answer to the wrong question.
Building your own library
These seven are a starting set, not a complete one. The prompts that will serve you best are the ones you write for your own recurring work.
1
Notice when you rewrite a prompt
The second time you type something similar is the signal. That is a candidate for saving.
2
Generalise it with placeholders
Replace the specifics with bracketed slots so it works next time without rewriting.
3
Store it somewhere findable in ten seconds
A single note titled "Prompts". If retrieving it takes longer than rewriting it, you will rewrite it.
4
Improve rather than replace
When a saved prompt underperforms, edit the saved version. Over months these compound into genuinely valuable assets.
Common mistakes
Pasting templates without filling the brackets properly
The placeholders carry the value. Vague fills produce vague output.
Collecting hundreds of prompts you never use
Ten prompts you use weekly beat two hundred you have bookmarked.
Using them on confidential material unchanged
Strip identifying and sensitive detail every time. Make it automatic.
Sending output without reading it properly
Your name is on it. Read every line as though you wrote it, because as far as the recipient is concerned, you did.
Your starting set
📝 Your prompt library
One place. Findable in ten seconds.
Write it down so it becomes automatic rather than a judgement call each time.
What comes next
These prompts are general-purpose. The next component gets specific — practical AI applications mapped directly onto seven professional roles, so you can see exactly what this looks like in your job.