How to Brief Your AI
You don't need perfect prompts. You need a clear task, a little context, and a simple way to improve what comes back.
What you’ll get
| The result | Your own reusable brief template, and a side-by-side test showing what a clear brief changes |
| Level | Level 1 (chat). Works in any AI assistant |
| Loop step you’ll practice | All of it: give context and instructions (Brief), adjust (Collaborate), and check (Check) |
| Time | About 10 minutes to read, 5 minutes to try |
1. Why talking to AI sometimes feels frustrating
Imagine you need to tell your team that a project will be delayed. You open ChatGPT, Claude, Copilot or another assistant and type:
Write an email about a project delay.
A few seconds later you have a professional-looking email. It’s polite, the grammar is fine, the structure looks good. But something is missing. It doesn’t explain the real reason for the delay. It may sound too formal for your team. It might even include details you never gave it. So you spend the next few minutes rewriting it.
Does that mean AI is bad at writing emails? Not really. It means you haven’t told it enough.
Think of AI as a smart, fast intern on their first day. They know a lot in general and nothing about your project. You wouldn’t hand a new intern a sticky note saying “write an email about a delay” and walk away. You’d explain what happened, who the email is for, and what you want people to do after reading it.
AI works better when you give it a useful brief, not just a vague request. That’s the first skill worth learning.
2. You don’t need a perfect prompt
You’ve probably heard the word prompt. It just means what you give an AI: a question, an instruction, a document, or a mix of these. “What is compound interest?” is a prompt. So is “Explain compound interest to a new colleague, using a simple example with $1,000.” Both are fine. The second gives more direction, but that doesn’t mean every request needs to be long.
If you want to know what a word means, a short question is enough. If you want help with an important email, a report or a decision, more context pays off. The goal isn’t the longest prompt. It’s giving AI what it needs for this task.
You also don’t need prompt formulas or phrases like “You are a world-class expert.” Ordinary language works. The major AI providers recommend the same basic habits: be clear about the task, give relevant context, describe the output you want, and refine as you go.
At Brief Your AI, we turn those habits into one simple method:
Brief → Collaborate → Check
If you’ve read The AI Map, you’ll recognize the loop: Brief is giving context and instructions, Collaborate is adjusting the draft together, and Check is you checking before you use it. Same loop, three easy words.
3. Step 1: Brief AI like a new intern
A brief is a short explanation of a task: what needs doing, plus what the other person needs to know to do it. A useful AI brief usually has four parts.
| Part | The question it answers | Example |
|---|---|---|
| Task | What do you want done? | Draft an email to my project team about a one-week delay. (Better than “Help me with my project.”) |
| Context | What does the intern need to know? | The testing milestone is delayed by one week because we need to fix two integration issues. The team already knows the project. |
| Output | What should the result look like? | A short email with a subject line and three paragraphs. |
| Boundaries | What should it avoid or respect? | Professional but friendly, under 150 words. Don’t invent a new delivery date. |
Context is where most of the improvement comes from: it’s what stops the intern from inventing a reason for the delay. But don’t add details just to make the prompt longer. Include what’s relevant. Boundaries matter most when facts, sensitive topics or business rules are involved.
Put the four parts together
Draft an email to my project team about a one-week testing delay.
We need the extra week to fix two integration issues. The team already knows the project.
Include a subject line, explain the delay, and ask team members to flag any dependencies.
Keep it professional but friendly, under 150 words. Don't invent a new delivery date.
This isn’t a complicated prompt. It’s a clear explanation of the work. And you don’t need all four parts every time: start with the task, then add context, output or boundaries when they’ll make a real difference.
A good brief isn’t about giving AI more words. It’s about giving it better direction.
4. Step 2: Collaborate
Even with a clear brief, the first answer won’t always be right. That’s normal. Treat it as a draft. Read it and ask yourself:
- Is this what I wanted?
- Is anything missing?
- Is the tone right?
- Has the AI assumed anything I didn’t say?
Then give feedback, in the same conversation. You don’t need to start over. If the email sounds too serious:
Make the tone more reassuring, but don't hide the fact that the project is delayed.
If it’s too long:
Shorten this to about 100 words. Keep the reason for the delay and the request for feedback.
Let the intern ask questions first
When you’re not sure what the AI needs to know, let it ask:
Before you write the email, ask me up to three questions about anything important that's missing.
This helps most with tasks involving several people, unclear requirements or important decisions. Don’t turn every simple task into an interview, though. If it’s straightforward, just ask.
Correct the direction, not just the wording
If you get a detailed technical explanation when you wanted something simple, don’t just say “try again.” Say what’s wrong:
Explain this to someone without a technical background. Use one everyday example and avoid jargon.
Don’t just accept or reject an AI answer. Improve it through conversation.
5. Step 3: Check what matters
An answer can sound confident and still be wrong. AI can produce wrong facts, misleading explanations, invented references and miscalculations. A clear brief reduces misunderstandings about your task; it can’t guarantee a correct answer.
This isn’t a theoretical risk. One study checked AI search answers sentence by sentence and found that roughly 25% to 47% of statements weren’t supported by the sources the tool itself cited (DeepTRACE, Salesforce AI Research, 2025).
How much you check depends on the stakes. Brainstorming names for a team event: low risk. A financial report, a contract or a health question: mistakes can be serious.
Three questions before you use an AI answer
- Are the facts right? Check important dates, numbers, names and claims against reliable sources. If the AI gives references, open them: don’t assume a citation exists or says what the AI claims.
- Has it invented or assumed anything? Look for details that weren’t in your brief. In the delay email, the AI might make up a new deadline or promise overtime. Those need correcting or confirming.
- Is it right for the purpose? A correct answer can still be wrong for the reader. A note to a close colleague should sound different from an announcement to a customer.
You’re still the one who decides whether it’s ready. For a fuller method, see How to Check Whether AI Is Telling the Truth.
A note about private information
Be careful what you share. Before uploading company documents, customer data or anything confidential, check your organization’s AI policy and the tool’s data settings. Not every AI service handles your information the same way.
Let AI help with the work. Keep control of the decisions that matter.
6. Three work examples to try
Brief → Collaborate → Check isn’t just for emails. Here it is on three everyday work tasks.
Example 1: Write a better email
Vague: “Write an email to my manager.” The intern has to guess what it’s about.
Brief:
Write a friendly but professional email to my manager asking to work from home next Friday.
I have a home appointment that requires me to be there. I can attend all meetings online.
Keep it under 100 words.
Collaborate: “Make it slightly less formal. Keep the explanation short.”
Check: Before sending, make sure the reason, the date and the working arrangement are right.
What you learned: a little context and a clear output turn a generic draft into something you can send.
Example 2: Understand a term everyone at work uses
Vague: “What is EBITDA?” You’ll get a textbook definition that may not help in your next meeting.
Brief:
Explain what EBITDA means. I work in operations and have no finance background.
Use a simple example with a small company's numbers, then tell me
why managers care about it and one common way it can be misleading.
Collaborate: If one part is still unclear, ask about just that part. Or turn it into practice: “Ask me three short questions to check I understand. Don’t show the answers until I try.”
Check: Compare the key definitions with a reliable source, such as your company’s finance team or an established accounting reference, before you use them in a meeting.
What you learned: AI can pitch an explanation at your level and help you learn through follow-up questions.
Example 3: Think through a work decision
Vague: “Which project management tool should we use?” The intern knows nothing about your team.
Brief:
Help me decide between two project management tools for my team: [Tool A] and [Tool B].
We're 8 people, mostly non-technical, and we use Microsoft 365.
What matters most: easy to learn, works with Teams, under $10 per person per month.
Compare the two on these points, explain the trade-offs,
and ask me any important questions before recommending one.
Collaborate: “Ease of learning matters more than features for us. How does that change your view?”
Check: Verify current prices, plan limits and integrations on each tool’s own website before you decide. These change often, and AI may be working from old information.
What you learned: AI can organize a decision, but the facts and priorities have to come from you.
7. Your first brief: try it now
Pick one real task you need to do today: a message, a topic you don’t quite understand, something to plan, ideas to organize. Open your AI assistant and start from this template. Fill in only the lines that matter:
Task: Help me [what I want done].
Context: Here's what you need to know: [relevant background].
Output: Please give me [the result or format I want].
Boundaries: Keep in mind [limits or preferences].
Then run the three steps: Brief (send it), Collaborate (ask for one specific improvement), Check (review it before you use it).
The five-minute challenge
Do the same task twice. First, a short, vague request. Then, in a fresh conversation, a clear brief with the context that matters. Compare the two answers: which one fits your needs better, and what still needs fixing?
You’ll probably find the clear brief saves time. You may also find that some tasks need very little extra explanation. Either way, you’ll learn what actually makes a difference for the work you do.
The one thing to remember
Talking to AI isn’t about secret commands or perfect prompts. It’s a communication skill: say what you want, give the context that matters, work with the answer, and check it before you rely on it. You don’t need to become a prompt engineer. You just need a better way to work with your intern.
Brief clearly. Collaborate thoughtfully. Check what matters.
Using a different tool?
This method works in any AI assistant, because it’s about what you write, not which buttons you press.
| Tool | Where to try it |
|---|---|
| Claude, ChatGPT, Gemini | Any normal chat |
| Microsoft 365 Copilot only | Copilot Chat, or the Copilot pane in Word or Outlook. Follow your company’s AI policy for what you can paste in |
Further reading
The habits in this guide match the prompting advice published by the major AI providers and the reliability guidance from standards bodies. The Brief → Collaborate → Check method is Brief Your AI’s own teaching framework, not an official standard.
- OpenAI: Prompting
- OpenAI: How do I create a good prompt?
- Anthropic: Prompt engineering overview
- Google: Prompt design strategies
- NIST: Generative AI Profile (AI 600-1)
- DeepTRACE study on AI search answers (Earth.com, 2025)
Next
Learn the Check step in depth: How to Check Whether AI Is Telling the Truth. Or put the method to work on a real task: start with Research Anything with AI (free), or try Let AI Interview You About Your Career.
Get new tutorials by email
One email when a new tutorial is out. No spam, unsubscribe any time.
Almost done: check your inbox and click the link in the email to confirm.
That didn't work. Check the address and try again.
I only use your address to send new tutorials. Privacy