Tools · Generator

Turn a vague AI idea into a buildable brief.

Nine short questions. A working draft brief in your screen in under two minutes. Then pressure-test it on a thirty-minute call.

Nine questions. About fifteen seconds to draft. No email needed.

01 · The problem

Most AI ideas die at the brief, not the build.

An AI idea that lives in a sentence almost never makes it into a shipping product. The reason is rarely the model. It's the brief.

Without a brief, the team optimises for the wrong thing. The conversation drifts between three different products. The cost-of-being-wrong is never named, so the system gets designed for the wrong stakes. The MVP balloons because nothing is out of scope on paper.

A brief is a forcing function. It moves the unknowns from the founder's head into a document people can disagree with on Tuesday.

02 · What a good brief includes

Six fields. No more, no less.

A workable AI product brief carries six fields. Most generic product templates have twenty. The ones below are the only ones that change a build decision.

  • WorkflowThe exact sequence of steps the work passes through today. Not what it should be. What it is. The brief stands or falls on this being honest.
  • User and contextWho runs the workflow, how often, and what they already use. Roles, not personas. Real tools, not categories.
  • AI roleExactly what AI is asked to do in the workflow and where the human stays in the loop. One sentence each. If you cannot finish the sentence, the role is not defined yet.
  • MVP scopeThe smallest version that proves the idea is worth carrying further. Not the smallest version that ships; the smallest version that teaches.
  • RisksWhat happens when the model is wrong, what data is touched, and what the failure mode looks like. Naming the worst case earns the right to ship the best one.
  • First build recommendationPrototype, sprint, MVP, or audit. Bound to a timeline. If you cannot name the next two weeks, the brief isn't done.

03 · Worked example

A brief for a fictional product.

What follows is a brief for a fictional idea: a tool that helps independent restaurants forecast tomorrow's ingredient demand from the last six weeks of sales. It is deliberately small. A brief that fits on a screen is a brief that lets a build start.

Product idea summary

A small forecasting tool for independent restaurants that turns the last six weeks of sales into tomorrow's ingredient order, sized for one supplier delivery cycle.

User and workflow

The head chef opens the tool on a phone at 9pm after service. The current workflow is a printed spreadsheet, last week's order, and gut. The tool replaces the last 20 minutes of that with a draft order the chef edits before sending.

AI role

Translates the last six weeks of POS data plus weather and reservation count into a quantity per ingredient. Stays out of pricing. Stays out of supplier choice. The chef approves before anything is sent.

MVP scope

One restaurant. Twelve top-selling ingredients. POS connector for one common system. A nightly job that drops the draft order into the chef's email by 8pm. No app, no dashboards, no team accounts.

Risks

Wrong forecast leads to waste or stock-out, both with real money behind them. The model must show its uncertainty so the chef adjusts. Data is POS only, no payments, no personal data. Failure mode is a manual order, which is the status quo, so the floor is sane.

First build recommendation

Two-week prototype with one restaurant. Goal is a single week where the chef approves the draft order without re-doing the math. If that week happens, fund the four-week MVP. If it does not, the brief was wrong and we go back to the workflow.

04 · How Axyom narrows it

Briefs come out of a conversation, not a form.

The most useful version of this tool is a thirty-minute call with the person who'll build it. The reason isn't gatekeeping. It's that the second question changes based on the answer to the first, and the third question depends on what surfaces in the second. A static form can't follow that thread.

An LLM-backed generator can get you most of the way to a draft brief. It will land in this page in a later phase, with the same six fields. Until then, the most reliable shortcut is the call.

[ Next ] · Start a project

Bring the idea. Leave with the brief.

Book a 30-minute call. We walk through the six fields, name the unknowns, and you leave with a draft brief in your inbox the same day.

Back to tools

Other tools

Related case study

Axyom