1. Build your prompt
Only the response experiment selects a fixture. No AI inference takes place.
THE WORKSPACE AUTOMATION SERIES · LAB 02
Turn a cell of text into a useful first draft. Learn every step between reading a sheet and reviewing an AI response.
16 lessons · 6 response experiments · 1 complete starter project
Keep the original. Send only what you need.
Check the status. Read the actual response.
Helpful automation starts with a review step.
A COACH AT EVERY STEP
CHANGE ONE THING. FOLLOW THE RESULT.
Only the response experiment selects a fixture. No AI inference takes place.
Build a request, then run the simulation.
No exchange yet.
B2 · Empty
This diagram models the exported script’s stages. Browser controls do not execute Google services.
FROM PRACTICE TO A REAL PROJECT
The starter makes a real API request when you run it in Google Apps Script. API access, authorization, availability and usage charges depend on your account. It processes A2 once and writes a draft to an empty B2 after confirmation.
Put non-sensitive notes in A2. Keep the starter’s summary task, generate once, then compare each claim to the original. Leave uncertain dates uncertain.
Change the task in buildPrompt_ to request a polite draft. Include approved policies in the prompt. Review promises and facts yourself. This starter does not send email.
Ask for owner, action and due date in plain text, using “unknown” when missing. If you later request JSON, validate its shape and every required field before writing columns.
Use explicit file IDs for background jobs, record row status, limit each batch and plan for retries. Avoid calling the API from a simple onEdit trigger or a custom formula that may recalculate repeatedly.
Loading downloadable starter…