Score trial users six hours after signup with an LLM and send a reviewed follow-up email (about $0.20 a day)

Trial users often sign up, look around and leave without reaching the first useful result, and a solo founder cannot manually research every signup.

Six hours after signup, a job asks an LLM (the author thinks GPT-4o) to explain and then score the user from 0 to 10 from email domain, name, team name, searches, pages visited and items viewed. The prompt favours the author's target customers and asks the model to check business domains. Scores of 5 or more get a fuller profile with suggested engagement, stored on the user record and posted to Slack. A dashboard lists trial prospects from the last 10 days with a button that drafts a follow-up from their activity, score and profile, usually recommending one next step such as setting up an alert; the founder reviews and sends each email himself. At signup, a small model (GPT-4o mini) turns the user's self-described role and project into three keyword groups for a first alert. In-app AI builders turn free text into alerts and podcast lists. All prompts share product documentation context that Claude condensed from a transcribed recording of a product walkthrough. User-triggered AI calls are rate limited and usage alarms are set.

The author says the follow-up email helped inactive users set up their first alert and that some barely-active prospects became paying customers a day or two after a suggestion. He gives no conversion rates, sample size or revenue figures. Cost is estimated at under a couple of cents per scored user and about five cents per emailed prospect, roughly 20 cents a day in total.

Self-reported by the product's owner with no numbers on conversion lift. Cost figures are estimates. The onboarding keyword feature had shipped only days earlier. Only about a fifth of users fill in the project description. Emails are sent manually, so it depends on the founder's time. The post does not discuss scoring accuracy, privacy or consent for profiling prospects.

A product with signup and activity data you can pass to an LLM, a Slack channel for alerts, and product documentation to use as prompt context.

• The Bootstrapped Founder (Arvid Kahl) ↗ Firsthand write-up, published August 29, 2025

Source published: August 29, 2025. Last reviewed here: October 15, 2026. Spot a mistake? Tell us.