A useful brief makes a commercial hypothesis testable. It does not pretend that a public signal proves buying intent.
Original Role Atlas portfolio exercise
At a glance
- Time to build
- 6-10 focused hours (estimate)
- What you need
- A spreadsheet and basic research skills. Optional AI assistance; no paid enrichment service or coding required.
What to ship
- A 20-company research sheet with source URLs and dates
- Five evidence-linked account briefs and a visible scoring rubric
- A manual-versus-assisted quality and time comparison
Choose a product and a narrow customer
For this original exercise, imagine selling an AI tool that helps small support teams find answers in their documentation. Define the customer you would test first, the problem you believe they have, and a reason the product might not fit.
Bessemer's demand-generation guide is optional background for thinking about customer selection. Your customer profile is a hypothesis, not evidence of demand.
Source / Bessemer: A founder's guide to generating demand (opens in a new tab)
Build the evidence sheet
Pick 20 companies with public websites. Record their product, relevant public documentation, an observed operational signal, its URL, and the date. Use permitted public sources; do not scrape private profiles or collect personal contact details.
- Keep separate columns for observed fact, inference, and unknown.
- Score fit using three visible criteria; do not hide a made-up confidence score behind AI.
- Draft five short briefs: likely need, supporting evidence, an open question, and a reason not to pursue.
- If you generate outreach examples, leave them as unsent drafts. No automated sending is needed for this project.
Compare quality, not volume
Time a manual brief and an assisted brief on comparable accounts. Check each factual claim against its cited source. Report unsupported-claim rate, missing-source rate, and time spent reviewing, not only generation time.
Have a peer review several briefs using the same rubric, if possible. Publish a sample sheet and a short decision memo. Do not claim pipeline or revenue impact unless you actually ran a permitted campaign and measured those outcomes.
How to describe it on your resume
Created an evidence-linked account research workflow for [N] companies; compared manual and AI-assisted briefs on [rubric], recording [measured quality/time result].
Use only after completing the work. Replace brackets with your own evidence. Label independent, prototype, and synthetic-data work accurately; do not invent customers or business impact.