# AI Brief Builder > A free AI prompt and brief builder for work: seven questions about one task, and a brief to paste into any AI tool, written from fixed wording, with no email required. Source page: https://mnpsystems.ca/tools/ai-brief/ MNP Systems is an engineering consultancy based in Brampton, Ontario, Canada, that takes work from anywhere: remotely, or on site where a job needs hands on the equipment. Contact: https://mnpsystems.ca/contact/, contact@mnpsystems.ca, or call 1-888-338-1090 (toll-free; an AI receptionist answers). **Ask AI for the bigger job, and stay the one who checks it.** Answer seven questions about one piece of work you are about to hand to AI and get a brief to paste into any AI tool: the outcome behind the task, the evidence and methods an expert would use, the boundaries and the reasons for them, and what will count as done. Then the questions to answer before you put your name to the result. It is built in your browser from fixed wording, so nothing you choose leaves this page unless you ask for the brief by email. You fill in the [brackets] with your own details after you paste it. The AI Brief Builder, walked through. A follow-up email for last month’s quotes briefed on an example: seven questions, one at a time, the brief written as they are answered, and the questions to answer before signing off. Video walkthrough (MP4, narrated): https://mnpsystems.ca/video/tools/ai-brief.mp4 It puts into practice the ideas in the post [How to Use AI at Work: Think Bigger, Check Harder](https://mnpsystems.ca/blog/think-bigger-with-ai/), which draws on Felix Rieseberg's "How to be more ambitious" and Addy Osmani's writing on intent debt, cognitive surrender and verification. ## The seven questions 1. **The work.** What are you about to hand to AI? (one of: A report, a deck or a review; An analysis of numbers or records; Writing that customers will read; A process the team repeats; A spreadsheet, a script or a small tool; Research for a decision) 2. **Altitude.** When it works, what changes? (one of: Someone makes a better decision; A problem is caught sooner; Customers act: they buy, reply or stay; The team gets hours back every week; Fewer mistakes reach customers) 3. **Depth.** What raw material can you give it? (any of: Source data: exports, records, spreadsheets; Earlier versions, and what happened after them; Customers’ own words: emails, calls, reviews; Specs, contracts or policies it must follow) 4. **Range.** Which methods should it borrow? (up to 3 of: An analyst’s: test every number; A journalist’s: find the story and lead with it; A skeptic’s: argue the case against; An editor’s: cut to what the reader needs; An engineer’s: find where it breaks; A teacher’s: explain it to a newcomer) 5. **Intent.** What must not change? (any of: The budget; The deadline; Our voice and brand; The tools and systems we already use; What we have promised customers; Legal, safety or compliance rules) 6. **Stakes.** If it is wrong and nobody notices, what happens? (one of: We redo it. Nobody outside the team sees it; A customer or the team acts on something wrong; Money, safety, someone’s job or a legal duty is on the line) 7. **Proof.** What will count as done? (one of: Every claim traces to a source we can open; It passes a test we set before it starts; The person who will use it says it works for them; A number we track moves once it is in use) ## An example brief For a quarterly review someone would otherwise ask for as "a deck", meant to help leaders decide, with the source data and last quarter’s review attached, an analyst’s and a skeptic’s methods, a fixed deadline and fixed systems, a medium cost of being wrong, and a test set before it starts: ### Before you send it - What I think the answer is: ____ - What success looks like: ____ ### The outcome (Altitude) - I need a report: [the task, in a line]. - It serves an outcome: to help [who decides] make a better decision about [what], sooner. Aim at that outcome, not only at the task. If a different piece of work would serve it better, say so before you start. ### The evidence and the method (Depth) - Work from the source data I attach, not from summaries of it. Tell me what is missing or looks wrong in it before you use it. - I am attaching earlier versions and what happened after them. Learn from what worked and what did not, and do not repeat what failed. - Work as a seasoned analyst would: compare against the same period before, separate one-offs from trends, and put what needs a decision first. ### Methods to borrow (Range) - Test every number: recompute each total and check each figure against its source. - Argue the case against your own answer, then say what survived and what did not. - Do each of these as a job with a result I can see, not as a voice to write in. ### The boundaries and the why (Intent) - Do not change the deadline and the tools and systems we already use. If a better result needs one of them to change, stop and tell me which one and why. - Why it matters: [the reason behind this work, and behind any boundary above]. - Where a choice depends on something I have not told you, ask rather than guess. ### What done means (Proof) - Done means it passes this test, which I have set before you start: [the test, e.g. run it on last quarter’s data and compare with what happened]. Run it and show me the results, including what the test does not cover. - Flag every fact, figure and promise that a customer or a colleague will act on. - When you finish, show me the evidence you used, the decisions you made that I did not ask for, and what remains unverified. ### Before you sign off - What did it decide that I did not ask it to? - Why do those choices work, in my own words? - What did the check prove, and what could still be wrong? - Check every fact, figure and promise someone will act on against its source. - Read it as if someone else wrote it. - Where could I rebuild the details if someone asked me? - One thing that was not checked: Fill in the [brackets] before you send it. Works in any AI tool: ChatGPT, Claude, Copilot, Gemini. ## Training the team Briefing AI and checking what comes back is part of the AI practice's Team training & AI policy service: hands-on sessions on the team's own work, and a short playbook for each role. ## What is kept Nothing. The brief is built in the browser. Only a visitor who asks for it by email sends their answers, and the brief is built again from them to write that email.