Case study
Lead qualification tool
A tool that screens prospect websites against fit criteria and sends only qualified companies to the sales team, already researched.
Built and used in a live outbound sales role.
- graded correctly at v1.3
- ~90%
- from start to v1.3
- 7 days
- cloud AI spend
- $0
- consumer GPU (RTX 2060)
- 8 GB
The problem
In outbound sales, much of the day goes to opening prospect websites one by one to decide whether a company is worth contacting. Most aren't, and that time is gone either way.
What we built
- A scraper visits each prospect's website and reads its content.
- Each site is searched for ideal-customer keywords and qualification terms, then graded against fit criteria by a language model running locally.
- Companies that pass the checks are uploaded to Apollo, enriched with a short AI-written summary of why each one qualifies, ready for outreach.
The result
Reps started from a pre-screened list with the research already done, instead of qualifying every website by hand.
Infrastructure and cost
- Runs on a desktop with an 8 GB RTX 2060, using Qwen 7B locally. No cloud AI APIs, so grading costs nothing per lead.
- Built in 7 days, from first line of code to version 1.3.