Most people who use AI at work type a fresh request every time. They think of a task, write a sentence or two, read the result, and then spend ten minutes fixing it. The next day they do the same thing again with a slightly different sentence and get a slightly different quality of answer. That is slow, and it makes the results feel random.
A prompt template fixes this. It is a reusable shape for a request, with blanks you fill in each time. In this guide you will learn what makes a template work, a set of ready-to-adapt templates for everyday tasks such as emails, summaries, planning and editing, how to fill them in well, and how to improve them over time so they get better instead of going stale.
Key takeaways
- A good template names the role, the task, the context, the format and the limits, with clear blanks for the parts that change.
- Templates save time and make quality consistent, but only if you fill the blanks with real detail instead of leaving them vague.
- Always give the AI the source material it needs, and tell it what to do when information is missing rather than guess.
- Test every template on a real task, note what went wrong, and edit the template, not just the single answer.
- Keep your best templates somewhere easy to reach, and never paste private or sensitive data into them without checking your tool’s rules.
Why templates beat fresh prompts
Every time you write a prompt from scratch you make a dozen small decisions: how much background to include, whether to ask for a particular length, what tone to request, what to forbid. Most of those decisions are the same from one week to the next. A template makes them once, carefully, and then lets you reuse them when you are tired, rushed or distracted.
Consistency is the main benefit
If you ask for a client update on Monday with a detailed prompt and on Thursday with a lazy one, the two updates will not match in quality or structure. A template keeps the structure steady. Colleagues who read your work start to expect a certain shape, and that makes review faster.
Templates also teach you what matters
When you build a template you are forced to decide what information the AI really needs. That habit carries over to every other prompt you write. If you want the wider skill behind it, our guide to writing better prompts covers the principles, and this article turns those principles into reusable pieces.
When a template is the wrong tool
Templates suit repeated tasks with a predictable shape. They are a poor fit for one-off, open-ended exploration, where a conversation that evolves is better. If you find yourself bending a template out of shape to fit a task, drop it and write a fresh prompt instead.
The anatomy of a good template
Nearly every useful template is built from the same handful of parts. You do not need all of them every time, but knowing the set lets you decide what to keep.
| Part | What it does | Example wording |
|---|---|---|
| Role | Sets the point of view and level of expertise | Act as an experienced editor for a general audience |
| Task | States the single thing you want done | Rewrite the text below so it is clearer and shorter |
| Context | Gives the background the AI cannot guess | The reader is a customer who is annoyed about a late delivery |
| Source material | Provides the text or facts to work from | Use only the notes between the markers below |
| Format | Describes the shape of the answer | Three short paragraphs, then a bulleted list of next steps |
| Limits | Sets length, tone and things to avoid | Under 150 words, plain English, no jargon |
| Fallback | Says what to do when something is missing | If a detail is not in the notes, write “not stated” instead of guessing |
Use clear blanks
Mark the parts that change with square brackets and a short description, such as [AUDIENCE] or [PASTE NOTES HERE]. Square brackets are easy to spot and easy to search for, so you will not accidentally send a template with a blank still in it. Make each label specific. [DETAILS] tells you nothing; [DATE, TIME AND LOCATION OF THE EVENT] tells you exactly what to add.
Separate instructions from material
When you paste long source text into a prompt, put it between clear markers so the AI can tell where your instructions end and the material begins. Something as simple as a line reading START OF NOTES and another reading END OF NOTES works. It also reduces the chance that the AI treats text inside your material as an instruction to follow.
Templates for writing and editing
Writing tasks are where most people start, and they respond well to templates because the shape of a good email, summary or announcement rarely changes.
Template: professional email
- Prompt: Write an email from [YOUR ROLE] to [RECIPIENT AND THEIR ROLE]. The purpose is [ONE-SENTENCE PURPOSE]. Key points to include: [LIST]. The tone should be [WARM / FORMAL / DIRECT]. Keep it under [NUMBER] words, with a clear subject line and one specific request at the end. Do not add facts that are not in the points above.
The last sentence matters. Without it, AI tools often add invented details to make an email sound complete. If you are working on marketing messages specifically, our guide to using AI for email marketing goes into more depth on audience and consistency.
Template: edit my draft
- Prompt: You are an editor. Improve the draft below for clarity and flow without changing its meaning or my voice. Fix grammar, cut repetition, and shorten long sentences. Then list the five most important changes you made and one thing I should reconsider. Draft: [PASTE DRAFT].
Asking for a list of changes turns the edit into a lesson, and it lets you spot any change you disagree with before you accept it.
Template: rewrite for a different audience
- Prompt: Rewrite the text below for [NEW AUDIENCE], who [WHAT THEY KNOW AND CARE ABOUT]. Keep every fact. Replace specialist terms with plain words or explain them briefly. Aim for a reading level suitable for [LEVEL]. Text: [PASTE TEXT].
Template: first draft from an outline
- Prompt: Turn the outline below into a first draft of about [LENGTH] words for [AUDIENCE]. Use short paragraphs and plain language. Where the outline is thin, add a clearly marked note saying what example or evidence I should supply, instead of inventing it. Outline: [PASTE OUTLINE].
That last instruction protects you from made-up examples, a problem we explain in our guide to AI hallucinations.
Templates for summarising and thinking
AI is good at compressing and restructuring text, which makes it useful for reading and analysis tasks. The risk is that a summary sounds confident even when it has missed something, so these templates build in checks.
Template: summarise a long document
- Prompt: Summarise the document below for [READER AND THEIR PURPOSE]. Give me: a three-sentence overview; the five most important points as bullets; any decisions, deadlines or numbers mentioned, copied exactly; and a list of questions the document leaves unanswered. Do not include anything that is not in the document. Document: [PASTE TEXT].
Template: extract action items
- Prompt: From the notes below, list every action item as a table with columns for the task, the owner and the deadline. If an owner or deadline is not stated, write “not stated”. Do not guess. Notes: [PASTE NOTES].
This pairs well with the habits in our guide to meeting notes with AI, where the quality of the raw notes decides the quality of the output.
Template: explain a concept simply
- Prompt: Explain [CONCEPT] to someone who is [BACKGROUND, E.G. NEW TO THE TOPIC]. Start with a one-sentence definition, then give a plain analogy, then a worked example, then three common misunderstandings. Tell me which parts you are less sure about.
Template: argue both sides
- Prompt: I am deciding whether to [DECISION]. Give the strongest honest case for doing it and the strongest honest case against, each in four bullets. Then list the three facts I would most need to check before choosing. Do not tell me what to do.
Templates for planning and organising
Planning prompts work best when they include your real constraints. A plan that ignores your time, budget and people is a fantasy, however neat it looks.
| Task | Details to put in the blanks | Format to request |
|---|---|---|
| Project plan | Goal, deadline, team size, known risks | Phases with tasks, owners and dependencies in a table |
| Weekly schedule | Fixed commitments, energy levels, priorities | Day-by-day list with time blocks |
| Event plan | Date, headcount, venue limits, budget band | Checklist grouped by weeks before the event |
| Study plan | Exam date, current level, hours available | Weekly targets with short review sessions |
| Meeting agenda | Purpose, attendees, time available, decisions needed | Timed agenda with a goal for each item |
Template: break a goal into steps
- Prompt: My goal is [GOAL] by [DATE]. I have [HOURS PER WEEK] and these constraints: [LIST]. Break the goal into milestones, then into tasks of no more than two hours each. Put them in a table with the order they should be done. Flag any step that depends on someone else, and any assumption you had to make.
For a more personal use of the same idea, see our guide to planning your week with AI.
Template: prepare for a meeting
- Prompt: I have a [LENGTH] meeting with [WHO] about [TOPIC]. My goal is [OUTCOME]. Here is what I know: [BACKGROUND]. Suggest an agenda, the five questions I should ask, and the objections I should be ready for. Keep it to one page.
Templates for difficult communication
Hard messages are where a template helps most, because stress makes people write badly. Use AI to produce a calm first draft, then edit it so it sounds like you.
Template: decline politely
- Prompt: Draft a short, kind message declining [REQUEST] from [PERSON]. The real reason is [REASON, OR “DO NOT GIVE A REASON”]. I want to keep the relationship good and offer [ALTERNATIVE IF ANY]. No more than [NUMBER] words, no excessive apologies.
Template: give feedback
- Prompt: Help me give feedback to [PERSON AND ROLE]. What went well: [LIST]. What needs to change: [LIST]. The outcome I want: [OUTCOME]. Write a short script that is specific and respectful, describes behaviour instead of character, and ends with a clear next step.
Template: customer apology
- Prompt: A customer is unhappy because [WHAT HAPPENED]. What we can actually offer is [REMEDY]. Write a reply that acknowledges the problem in the first sentence, explains what we are doing, and does not promise anything beyond the remedy listed. Tone: [TONE].
Be careful with promises. The AI does not know your policies, so the sentence about not promising beyond the listed remedy is essential.
How to fill in the blanks well
A template is only as good as what you put in it. Three habits make the biggest difference.
Be specific about the audience
“Customers” is weak. “First-time customers who bought a subscription last week and have not logged in” is strong. The more precisely you describe the reader, the better the AI can choose vocabulary, length and emphasis.
Provide the facts yourself
If the answer depends on a date, a figure, a policy or a name, put it in. Never ask the AI to recall facts you could supply, because it may produce something plausible but wrong. After you get a result, follow the steps in how to check AI output for mistakes before anything goes out under your name.
Say what good looks like
If you have a sample you like, describe its qualities: short sentences, no exclamation marks, starts with the answer. Describing the target is more reliable than hoping the AI guesses your taste.
A before and after example
| Weak fill | Strong fill |
|---|---|
| [AUDIENCE]: staff | [AUDIENCE]: warehouse staff who read updates on their phones during breaks |
| [GOAL]: improve things | [GOAL]: reduce the number of late-day order changes by agreeing a clear cut-off time |
| [TONE]: professional | [TONE]: friendly but direct, no corporate phrases |
| [NOTES]: see attached | [NOTES]: pasted text between START and END markers |
How to improve a template over time
A template is a draft that you refine. Treat the first version as a hypothesis and the first few uses as tests.
- Use it on a real task. Real work exposes gaps that invented examples hide.
- Note what you had to fix. If you keep rewriting the opening, the template needs an instruction about openings.
- Edit the template, not only the answer. Add the missing rule or constraint to the template itself.
- Test again on a different task. A good template works across several inputs, not just the one it was built on.
- Record a version note. Write the date and what you changed so you can undo a change that made things worse.
Add examples carefully
Showing one or two examples of a good output can improve a template a lot. Keep examples short and generic, and tell the AI to match the style but not copy the content. If every output starts to look like your example, the example is too dominant and you should shorten or vary it.
Remove what does not help
Long templates are not better templates. If a line has never changed an output, delete it. A tight template of eight lines usually beats a sprawling one of thirty.
Common template mistakes
| Mistake | What goes wrong | Fix |
|---|---|---|
| Leaving brackets in place | The AI treats the placeholder as real text | Search for the opening bracket before sending |
| Packing several tasks in one | Each part gets less attention | Split into separate prompts or numbered stages |
| No fallback instruction | The AI invents missing facts | Add “if not stated, say so” |
| Pasting private data casually | Sensitive information leaves your control | Check your tool’s data rules first |
| Never updating | The template drifts away from your needs | Review your top templates every few months |
A note on sensitive information
Templates make pasting easy, which also makes oversharing easy. Before you build a template that takes customer details, contracts or health information, read our guide to protecting company data when using AI tools and replace real names and figures with placeholders where you can.
Template checklist
Run through this list whenever you write or revise a template.
- Does it state one clear task in the first lines?
- Does it describe the audience and purpose, not just the topic?
- Are all changing parts marked with specific bracketed labels?
- Does it separate instructions from pasted material with markers?
- Does it set a format, a length and any tone limits?
- Does it tell the AI what to do when information is missing?
- Have you tested it on at least two different real tasks?
- Is it stored somewhere you will actually find it, with a short name?
- Have you removed lines that never change the result?
- Have you checked it contains no private data baked into the wording?
A reusable master skeleton
If you are unsure where to start, copy this skeleton and adapt it: “Act as [ROLE]. Your task is to [TASK]. The audience is [AUDIENCE] and the purpose is [PURPOSE]. Use only the material between START and END. Format the answer as [FORMAT], in [TONE] language, within [LENGTH]. If anything needed is missing, list it instead of guessing. START [MATERIAL] END.”
Frequently asked questions
How long should a prompt template be?
As long as it needs to be and no longer. Most good templates run from five to twelve lines. If you cannot explain why a line is there, remove it and see whether the output gets worse.
Do templates work the same in every AI tool?
The general structure travels well, but tools respond a little differently, so test a template in each tool you use. A template that gives tidy results in one may need a stricter format instruction in another.
Should I tell the AI to act as an expert?
A role can help set tone and focus, but it does not make the AI more accurate. Specific context, real source material and clear limits do more for quality than a grand job title.
Where should I store my templates?
Anywhere you can reach in two clicks: a notes app, a shared document or a text expander. The best storage is the one you will actually open. Give each template a plain name that says what it does.
Can I use the same template for different people on my team?
Yes, and it is a good way to share good practice, but agree who maintains it so it does not fork into several slightly different versions. Our guide to writing an AI use policy for your team explains how to set simple shared rules.
Why does the output still come out generic?
Usually the blanks were filled with vague words. Add concrete details, real examples from your work and a clear description of what you want to avoid. Generic input produces generic output.
Conclusion
Prompt templates turn a scattered habit into a repeatable skill. You decide once what the AI needs to know, you mark the parts that change, and you keep improving the wording based on real results. The payoff is steadier quality, less time spent rewriting, and a shared way of working if others use the same tools.
Start small. Pick the three tasks you repeat most, build a template for each, and test them this week. When you are ready to go further, our guide to building your first AI workflow without code shows how to chain steps together, and how to choose an AI tool can help you decide where your templates should live.
