Who hurts
Solo founders and consultants who run the whole business out of one inbox.
The problem
The cost isn't reading email, it's the context switch. Ten threads a day each need thirty seconds of judgement and two minutes of writing, and they arrive scattered across the day.
What you build
An agent that reads overnight mail, classifies each thread against the user's own past replies, drafts responses in their voice as Gmail drafts, and leaves a single morning digest: four drafts to approve, everything else already filed.
Why now
Long-context models can hold a year of a person's sent mail as style reference, and Gmail's API lets you write drafts without ever sending — the risk profile that killed earlier attempts is gone.
Validate it this week
Run it manually for five people for a week. You read their mail, you write the drafts. If they don't approve at least half your drafts unedited, the product can't work either.
Why you'd keep winning
Voice fidelity compounds per user — every approved or edited draft is a training example nobody else has. Switching costs grow quietly.
The honest risk
One embarrassing auto-sent reply ends the account. Stay draft-only far longer than feels necessary, and expect that to cap perceived value.
The prompt is written to make an AI argue with you before it writes code — that first round of pushback is worth more than the scaffold.