Find the use case worth building
Turn a customer's vague wish for AI into one use case with a number that says when it worked.
You build: A scored list of five candidate use cases and a one-page brief for the winner.
- A real or imagined customer (Parcel Desk works)
- No code
-
01Write up three short interviews: who does what today, how long it takes, what goes wrong
ls interviews | wc -lYou pass when: Three files, each with at least one number someone gave you (hours a week, claims a day, error rate).
-
02List five candidate use cases and score each: reach, impact, confidence, effort
The top score is rarely the most exciting idea. It is usually the dull task many people do every day.
grep -c "^|" rice.mdYou pass when: A table with five scored rows and a total (reach times impact times confidence, divided by effort).
-
03For the winner, write the success number and how it will be measured
grep -n -i "success" BRIEF.mdYou pass when: A sentence with a number and a date, such as: 70 percent of claims under 100 decided with no human, at under 2 percent wrong payouts, by week six.
-
04Finish the brief: problem, users, success number, what is out of scope, what needs a human
wc -w BRIEF.mdYou pass when: Under 400 words. A person who was not in the interviews can say what you are building and how you will know it worked.
The check runs on the computer, not on your word: a step passes when its command exits with 0 there.
Computers are not set up on this site yet. Run each command on your own machine and compare with "You pass when".
Commands run in labs/50-discovery and may take up to 60 seconds each. Every command starts a new shell: put keys and settings in a .env file there, it is loaded each time.