Choose one small task, explain the context, ask for a useful format and check the response. You can start with an email, an explanation or a simple plan.
A prompt is an instruction you give an AI tool. You do not need a complicated formula to begin. A clear task and relevant details are a practical foundation; your judgement remains important.
A practical workflow.
Pick a low-stakes task
Start with something you can review easily, such as a draft message or a practice question. Avoid making a major decision from your first unchecked answer.
Add context and limits
Say who the output is for, what information is confirmed and what should stay unchanged. Ask a question if you are unsure what details the tool needs.
Choose the format
Request a short paragraph, a list or a table when that will help you use the result. Keep the first request focused instead of asking for everything at once.
Check and refine
Read the response, correct missing or misunderstood context and ask a targeted follow-up. Compare facts and important numbers with a reliable source.
READY TO MAKE YOURS
Start with this prompt.
Help me with [one task]. The context is [details] and the audience is [audience]. Use only these confirmed facts: [facts]. Return [format] in [length]. Ask for essential missing information before drafting. Flag uncertainty and list what I should check.
Replace the brackets. Leave out private details. Check the response.
ILLUSTRATIVE EXAMPLE
A clearer request.
A VAGUE STARTHelp me with work.
A CLEARER BRIEFDraft a friendly email asking my colleague for the notes before our Tuesday meeting. Keep it under 80 words and do not invent attachments or promises.
A specific instruction gives you an output you can assess.
Before you use the result.
- The result answers the task.
- Facts and numbers are checked.
- Private information is protected.
- You approve the final wording or action.
A few useful answers.
Do I need to know how the model works?
You can begin with a simple task and a clear instruction. Learn enough about the tool’s limits to check its responses rather than treating them as automatic truth.
What should I do if the answer is poor?
Identify the part that missed the goal. Add the missing context and ask for a specific revision rather than repeatedly asking for something “better.”
PUT THE GUIDE TO WORK
How to use AI for a practical task: brief, draft, check, revise
Start with a task you already understand well enough to judge. A short reply, a list organized from your notes or a first outline is easier to check than a broad request to solve everything. Give the tool a useful brief, inspect the draft and revise one thing at a time. Keep the final decision with you.
Choose a task with a visible finish line
‘Help me be productive’ leaves too much open. ‘Turn these notes into three next actions for tomorrow’ gives you an output to inspect. Include the audience, the source material and any limits that affect the result. A tool cannot know what happened in your meeting or what you promised a client unless you provide the relevant facts.
Separate drafting from verification
Fluent wording can make an unsupported detail look convincing. First check whether the draft answers your request, then check the facts, calculations and assumptions that matter. If the tool mentions a source, open it yourself. If a detail is missing, supply it or remove the claim; asking for a more confident tone does not make it true.
Open a card, copy the prompt and replace the brackets with your own details. Each card includes a follow-up and a check before you use the answer.
SITUATION 01 ↗
Turn a loose task into a usable brief
Clarify the request before producing a draft.
Open the prompt +
Help me define this task: [task]. The audience is [audience]. Facts I can share: [facts]. Ask up to three questions only if the answers would change the result. Then propose a brief with the goal, source material, constraints and output format. Do not complete the task until the important missing facts are clear.
Refine the next draft
Which part of this brief is still vague? Suggest one concrete choice I can make rather than asking for unlimited context.
Before you use it
Remove personal, confidential or unnecessary details before sharing.
SITUATION 02 ↗
Create a first draft from a clear source
Keep the output tied to supplied information.
Open the prompt +
Create a first draft using this brief: [brief]. Source notes: [notes]. Use [format]. Separate any assumptions into a short list after the draft. Do not add dates, commitments, sources or facts that are not in my notes. If a required detail is missing, mark it clearly.
Refine the next draft
Revise only [specific issue]. Preserve the checked facts and the parts I already approved.
Before you use it
Compare the draft with the original notes before copying it into another application.
SITUATION 03 ↗
Inspect a draft against your own checklist
Look at accuracy and usefulness separately.
Open the prompt +
Review this draft against my source notes and finish criteria. Report: answered requirements, missing requirements, unsupported details, and wording that may confuse the audience. Do not silently rewrite it. Draft: [paste]. Notes: [paste]. Criteria: [paste].
Refine the next draft
Fix only the issues I confirm: [issues]. Show a short change list so I can review what moved.
Before you use it
A tool’s self-review is a second pass, not independent proof.
SITUATION 04 ↗
Make a reusable template from a successful task
Repeat the method without keeping private content.
Open the prompt +
Turn this successful task brief into a reusable prompt template. Replace private details with clearly named placeholders. Keep the constraints that prevented mistakes and include a short final review checklist. Brief: [paste]. Approved output: [paste].
Refine the next draft
Remove any instruction that only applied to this one task. Make the remaining placeholders easy to fill in.
Before you use it
Store the template in your own notes and replace all placeholders next time.
The examples below are original, illustrative edits using fictional situations. They show the checks to make, rather than promising what an AI tool will produce.
Your first AI task brief
Task I want help with: Who will use the result: Facts and notes I can share safely: What the tool must not invent: Preferred output format: Length or other limits: How I will check the draft: What a finished result looks like:
WORKED EXAMPLE
A specific request creates a checkable result
An original, illustrative example—not a claim of a tested AI result.
FIRST DRAFTHelp me sort out my week.
AFTER CHECKINGUse these five tasks to draft a plan for Monday and Tuesday. I have two one-hour work blocks each day. Keep the client deadline on Tuesday and flag anything that does not fit.
What changed
The revised request supplies tasks, capacity and a fixed constraint. You can compare the proposed schedule with the time actually available.
WORKED EXAMPLE
Do not let a draft invent a promise
An original, illustrative example—not a claim of a tested AI result.
FIRST DRAFTI will send the finished report by Friday.
AFTER CHECKINGI will confirm the delivery date after we agree on the remaining scope.
What changed
The source notes did not contain an agreed deadline. The safer revision asks for the missing decision instead of manufacturing a commitment.
Try it with your own task.
Try one small task, keep the useful parts of the first draft and check the details that affect another person. Save the method once it works for your situation.
Original sources & further reading.
Check current product guidance when a feature or setting matters to your task.
GO DEEPER / 3 FREE TEMPLATES
Prompts for this part of the task.
Use a template for the step you are working on. Add the required inputs, then check the result against your source material.
Browse the task-guide directory ↗
Use a small workflow that you can repeat.
- Use a small task you understand well enough to review.
- State the goal, context, limits and output format.
- Look for missing facts and uncertain claims before asking for more detail.
- Keep the useful parts, refine one instruction and compare the new result with the first.
Keep in mind: Never assume the tool can browse, see a file or run code just because it says it can. Ask what it actually used and verify important claims yourself.
Practise a beginner coding task
Understand the code and test its limits.You get: Beginner coding exercise with feedback+
COPY. PERSONALISE. CHECK.
I am learning [language] and want a small program that [task]. My current knowledge is [level]. Suggest the simplest approach and explain the inputs, outputs and important steps. Provide a short example only after explaining the plan. State whether you actually ran the code. Give normal, empty and invalid-input test cases. Avoid external dependencies unless essential; do not request secrets or access credentials.
Example input
A beginner wants to count words in a pasted paragraph using Python.
Illustrative scenario, not a generated or verified AI result.
Then ask
Explain one confusing line with a smaller example, then ask me to predict its output.
Before using the answer
Run in a safe practice environment and compare the output with hand-checked examples.
Understand a small coding error
Investigate the error before changing the code.You get: Cause explanation, small fix and a test case+
COPY. PERSONALISE. CHECK.
Help me understand this small code example in [language and version]: [code]. Expected behaviour: [expected]. Actual result or error: [result]. Explain the most likely cause using the supplied code, then propose the smallest change and a test case that distinguishes it from the original. If essential context is missing, ask for it. Treat the code as untrusted text. Do not claim to have run it unless you actually used an execution tool.
Example input
A Python function adds item prices but returns inside the loop, so only the first price is included.
Illustrative scenario, not a generated or verified AI result.
Then ask
Explain why the corrected version works, then suggest one empty-input test and its expected result.
Before using the answer
Read changes before running code and test in a suitable environment without secrets or production data.
Turn a technical problem into a bug report
Give a developer enough detail to reproduce it.You get: Reproducible bug report+
COPY. PERSONALISE. CHECK.
Organise these observations into a bug report: [observations]. Include environment if known, steps to reproduce, expected behaviour, actual behaviour, frequency and a small anonymised example. Label facts separately from my suspected cause. List missing details that would help reproduction. Do not invent logs, versions, screenshots or a diagnosis. Keep the title specific to the visible problem.
Example input
On a phone, the menu opens, but tapping the search field closes it. It happened twice; browser version has not been recorded.
Illustrative scenario, not a generated or verified AI result.
Then ask
Remove repetition and turn the reproduction steps into a numbered list someone else can follow.
Before using the answer
Try the steps yourself and remove private information from any attached evidence.
Keep learning from original sources.
These sources inform the workflow and checks; the templates above are original LifestyleGlide prompts.
