Live Dec 7 & 14 · Free founding cohort Give your idea two Mondays ↗
The Builder’s Journal
BUILD WITH AI · 3 MIN READ

Use ChatGPT to turn a product idea into a better brief

A discovery workflow for organizing evidence, narrowing scope, drafting requirements, and planning a useful first test.

Give the work a home

ChatGPT projects can keep related chats, files, instructions, and sources together. That makes a project useful when your product work spans research, a brief, messaging, and later revisions. Upload or connect only the sources you want the assistant to use; a project does not automatically know everything about your business.

Create a simple source pack: the problem you observed, anonymized interview notes, the current workaround, your constraints, and a list of open questions. Date the notes so old assumptions do not quietly become current facts.

Ask for evidence, assumptions, and gaps

Using only these notes, separate observed customer behavior, direct statements, my assumptions, and unanswered questions. Link each theme to its source. Identify contradictory evidence. Do not invent interviews, quotes, market size, or demand.

Check the result against the original notes. If a theme is based on one comment, keep that limitation visible. AI is useful for organizing material, but an organized document can still be wrong.

Explore three ways to deliver the same outcome

Before deciding to build software, ask for a manual service, a lightweight resource, and a software option for the same customer job. Compare the effort to create each, the effort to deliver it, and what it would teach you.

For this audience and problem, propose three small formats. For each, explain the useful result, the smallest test, the major assumption, and what would make it a poor fit. Use my available time and budget as constraints.

You choose the format. The exercise helps expose tradeoffs; it is not a forecast of which product will sell.

Write a brief a builder can act on

Your brief should name the audience, triggering situation, outcome, input, main steps, output, and exclusions. Add acceptance conditions that describe observable behavior. For a planning tool, what must a user be able to finish, edit, save, or export?

Ask ChatGPT to point out ambiguity and hidden dependencies. If it proposes accounts, payments, or automation, ask which user need requires each one in the first version. Move optional features to a later list.

Prepare the test before you polish the pitch

Create a task someone can attempt without your narration. Define what you will observe: completion, confusion, corrections, or a request to use the result again. Draft neutral follow-up questions about their experience.

For example, ask someone to make a plan for a real project using your worksheet. Watch which fields they skip. A suggestion to rename a field is more useful when you know the misunderstanding behind it.

Keep a decision log

After each test, record what changed, the evidence behind the change, and what remains uncertain. Ask ChatGPT to update the brief using the approved decisions and highlight any conflicts with earlier instructions.

Use it to prepare clear handoffs to Codex, Claude Code, or a human collaborator. These roles can overlap; the purpose of the handoff is to preserve the product decisions as you move into implementation.

Your first action: gather one page of real notes and run the evidence-sorting exercise. You will get more useful direction from that than from repeatedly asking whether an idea is good.

Make your next step a real one.

Use a free worksheet to define your product, or explore the Two Mondays workshop for time to build with Ashley.