What this toolkit is: A ready-to-use prompt library for your organisation — structured templates organised by role and task type, with fill-in-the-blank fields your team customises once and reuses forever. This is your floor. Prompt engineering training builds the ceiling.
Day 9 gave leadership the strategy framework — problem clarity, ownership, and success definition. Today we shift from the boardroom to the team. The most common reason AI adoption stalls at the individual level isn't resistance. It's that people don't know how to ask well. This toolkit solves that with a prompt library your organisation can deploy on Day 1.
In most organisations today, the quality of AI output is less about expertise in a domain and more about how naturally someone constructs a sentence in English. That's not a capability gap — it's a language gap wearing a capability gap's coat. The fix isn't more training, not yet. The fix is turning prompting from an art into a science first. A structured prompt library gives every team member the same starting point regardless of how they express themselves naturally. Once the floor exists, training improves from something that works — not from nothing.
Your teams need a floor before they can build a ceiling.
Master Template
You are a [your role — e.g. Project Manager / HR Business Partner / Finance Analyst] at [company name or type — e.g. a mid-size logistics company], working in [department].
I need to [specific task description — be as precise as possible].
The audience for this output is [who will read or use this — e.g. the CEO / my team / a client / all staff].
The tone should be [formal / conversational / technical / executive-level / simple and clear].
A good output looks like [describe the format — e.g. 3 short paragraphs / a bullet list / a table with 3 columns].
Do not include [anything to exclude — e.g. jargon / personal opinions / more than 200 words].
You don't need every line every time. Start with role, task, and audience — those three alone will dramatically improve your results compared to a bare question. Add the other lines when you need more precision.
Use before any performance, conduct, or sensitive 1:1 conversation. Fill in the issue precisely — vague inputs give vague outputs.
Prompt
I am a [role] at [company]. I need to have a conversation with a team member about [specific issue — e.g. three consecutive missed deadlines / a conduct concern raised by a colleague / consistent underperformance in Q2].
Help me structure this conversation. The outcome I want is [e.g. a clear and agreed improvement plan / mutual understanding of the issue / a documented verbal warning].
Tone: firm but fair. Give me an opening line, three key points to cover, and a closing that confirms next steps.
Hint: The outcome field is the most important. "I want to be understood" and "I want a signed improvement plan" require completely different conversations. Know which one you need before you walk in.
The "no jargon" instruction matters — AI mirrors the jargon in your inputs unless you push back explicitly.
Prompt
I am a [role] at [company]. Write a concise status update for [project or initiative name] to be shared with [audience — e.g. the C-suite / the steering committee / the board].
This period: [what happened — key progress, completions, decisions made].
Risks or blockers: [any concerns, delays, or issues].
Next steps: [what is coming in the next period].
Format: 3 short paragraphs — This Period / Risks / Next Steps. No jargon. No passive voice. Under 200 words.
Hint: Paste your raw bullet notes into the three fields. AI structures them into a readable update. You review and adjust the facts — not the writing.
Works with rough notes, bullet points, or voice-to-text transcripts. AI handles messy input well when the output structure is clear.
Prompt
Below are my notes from a meeting at [company] about [topic]. Attendees included: [list roles only — e.g. CTO, two senior engineers, Head of Product].
Summarise the key decisions made, list action items with role-based owners, and flag any unresolved questions.
Format — four sections:
— Summary (3 sentences max)
— Decisions Made
— Action Items (owner role | action | deadline if mentioned)
— Open Questions
Here are my notes:
[Paste your raw notes here]
Hint: List roles not names in the attendees field. It keeps the prompt reusable and avoids including personally identifiable information in your AI tool.
Specifying the most important outcome forces the agenda to have a purpose, not just a list of topics.
Prompt
I am a [role] at [company]. Create a [duration — e.g. 60-minute] meeting agenda for [meeting purpose — e.g. Q3 planning / project kickoff / retrospective].
Attendees: [roles in the room].
The single most important outcome of this meeting: [one decision / one alignment / one commitment].
Topics to cover: [list your topics].
Include time allocations. Mark one item as the decision point. End with a 5-minute close covering next steps and owners.
Hint: If you can't fill in "the single most important outcome" — consider whether the meeting is needed at all.
The analogy instruction is the most important part. It forces the model to translate, not just simplify.
Prompt
I am a [technical role — e.g. Cloud Architect / Security Engineer / IT Manager] at [company]. Explain [technical concept — e.g. API gateway / zero trust architecture / Terraform / cloud migration] to a [non-technical audience — e.g. CFO / HR Director / Board member] who has no technical background.
Use one everyday analogy to explain the core idea. Under 150 words. No acronyms unless you explain each one immediately. End with one sentence on why this matters to the business.
Hint: If the first analogy doesn't land, ask: "Give me three alternative analogies for the same concept." Pick the one that fits your specific audience.
The "facts only" field is critical — never let AI fill in what it doesn't know in a security communication. Review carefully before sending.
Prompt
I am a [role — e.g. CISO / IT Manager / Head of Technology] at [company]. Draft a communication to [audience — e.g. all staff / affected customers / the board] about a [type of incident — e.g. phishing attempt / system outage / data incident].
What we know (facts only — no speculation): [confirmed facts only].
What we are doing: [specific actions taken or underway].
What we need from the audience: [specific ask — e.g. change your password now / no action required / report suspicious emails to IT].
Tone: calm, clear, no blame, no jargon. One clear call to action at the end.
Hint: Leave the "what we know" field brief if you don't have full information yet. A communication that acknowledges uncertainty clearly is better than one that speculates.
Works well with raw notes, email threads, or meeting transcripts. Let AI impose structure on your thinking, then review for accuracy.
Prompt
I am a [role] at [company]. Organise the following raw notes into a structured requirements summary for [system or project name].
Five sections:
1. Overview (2 sentences)
2. Functional Requirements (what it must do)
3. Non-Functional Requirements (performance, security, availability)
4. Constraints (budget, timeline, technology, compliance)
5. Assumptions (what we are taking as given)
Do not add requirements not in the notes. Flag anything incomplete with [NEEDS CLARIFICATION].
Here are the raw notes:
[Paste your notes here]
Hint: "Do not add requirements not in the notes" prevents AI from hallucinating plausible-sounding requirements. Always review output against your source material.
The output gives you the right questions to ask — not just generic advice about "right-sizing."
Prompt
I am a [role] at [company]. We run workloads on [cloud provider — e.g. AWS / Azure / GCP]. Approximate monthly spend: [rough amount or range]. Key services: [e.g. EC2, RDS, S3, Lambda, EKS].
Give me:
1. The five areas most likely to have quick wins for my service profile
2. Three questions to ask my cloud team before the review
3. Three questions to ask the cloud vendor during the review
4. One metric most teams ignore but should track monthly
No generic advice — be specific to the services I listed.
Hint: "No generic advice — be specific to the services I listed" forces AI to be useful rather than broadly accurate. Apply this pattern to any domain-specific prompt.
"Keep my meaning exactly" is the single most important constraint. Without it, AI tends to rewrite rather than refine.
Prompt
I work as a [role] at [company]. Improve the clarity and flow of the following text.
Rules:
— Keep my meaning exactly. Do not add new ideas or remove existing ones.
— Audience: [who will read this].
— Make it [shorter / more formal / easier to understand / more direct / punchier].
— Maximum length: [word count if relevant].
Here is my text:
[Paste your text here]
Hint: If the output drifts from your meaning, say: "You changed [X]. Restore my original meaning while keeping the improved clarity." Iterating in the same conversation is faster than starting over.
The "five questions to ask" instruction is the most valuable part. It tells you what you don't know you don't know.
Prompt
I am a [role] at [company]. I need to quickly understand [topic] well enough to [what you'll do with it — e.g. present to leadership / make a decision / have a credible conversation with a vendor].
Give me:
1. A plain-language explanation in 3 to 4 sentences — assume I know nothing
2. The 3 most important things to know
3. The most common misconception about this topic
4. Five questions I should be asking — including ones I wouldn't think to ask yet
No jargon without immediately explaining it.
Hint: The "what you'll do with it" field changes the output significantly. AI calibrates depth and framing to the purpose. Always tell it where you're going, not just where you are.
The "main message in one sentence" forces you to clarify your own thinking before asking AI to write. Often the most useful step.
Prompt
I am a [role] at [company]. Write a professional email to a [recipient role — e.g. vendor / colleague / client].
Subject: [what the email is about in one line].
My main message in one sentence: [the single thing you need to say or ask].
Context they need: [any background the recipient needs].
Tone: [formal / friendly / firm / apologetic / direct].
Length: [short under 100 words / medium 150–200 words / detailed].
Call to action: [what you want them to do — e.g. confirm receipt / schedule a call / approve by Friday].
Hint: If you can't fill in "main message in one sentence" — the email isn't ready to write yet. That difficulty is the prompt working as intended.
"No filler" prevents AI from padding with transition phrases that add length but remove clarity.
Prompt
I am a [role] at [company]. Turn the following document into a one-page decision brief for [audience — e.g. the CEO / the steering committee].
Five sections only:
1. Context — 2 sentences maximum
2. The decision required — one clear sentence
3. Options — maximum 3, one line each
4. Recommendation — one option with one reason
5. Key risks — maximum 3 bullets
Direct. No filler. No passive voice. Do not add information not in the original document.
Here is the document:
[Paste document here]
Hint: "Do not add information not in the original document" prevents AI from supplementing your brief with plausible-sounding additions that aren't yours to make.
Apply these to every template — and every prompt you write from scratch
1
Role first, always
Starting with "I am a [role] at [company]" sets context before AI reads anything else. It changes the register, depth, and framing of the entire response.
2
Specify what good looks like
"A good output looks like X" is more useful than "make it good." Describe the format, length, sections, or give an example. Concrete beats abstract every time.
3
Tell it what to exclude
"Do not include [X]" is as important as telling it what to include. AI fills gaps by default — with jargon, caveats, generic filler. Exclusions are constraints, and constraints improve output.
4
Iterate, don't restart
If the first output is 70% right, say "keep the structure, change the tone to X." Build on what worked. Refinement in the same conversation is almost always faster than starting over.
5
Shorter prompts, more rounds
A clear 4-line prompt outperforms a 20-line prompt trying to anticipate everything. Ask, review, refine. You learn what the model needs by seeing what it produces.
6
Save what works
When a prompt gives you a great result, save the filled-in version. That's your personal library growing in real time — and the foundation for sharing what works across your team.
⚠️
A Note on Privacy and Data
- Use general descriptions rather than specific confidential data. "A mid-size logistics company" is more appropriate than your actual company name in most AI tools.
- Use roles not names when describing people — especially in performance or HR templates. This applies to all sections of this library.
- Check your organisation's AI usage policy before pasting internal documents, client data, or commercially sensitive material into any AI tool. When in doubt, anonymise first.
Where This Leads
These templates are the floor — they give every person on your team the same starting point regardless of how naturally they express themselves in English. Once teams are getting consistent results from these starting points, they're ready for the next level: prompt engineering training that teaches chain-of-thought prompting, few-shot examples, and role-playing techniques.
Build the library first. Train upward from something that works. The series continues at Day 11.