A prospecting pipeline that thinks
Built at Ruvo, 2026
Thousands of companies open in Brazil every month, and some of them are perfect customers for a cross-border fintech. Finding them by hand took about 30 minutes per company. I built a pipeline that does it in seconds.
How it works
- FindA Python script pulls newly opened companies from public data and gets their registration details from a government API.
- EnrichAI agents with MCP tools enrich each company with third-party data and sort it by sector and niche.
- ScoreAnother script rates how well each company fits our ideal customer, from 0 to 10.
- WriteAI writes a personal first message using the value proposition that fits that company best.
- SendThe list connects to WhatsApp and email, so a person reviews and sends with one click.
Decisions that mattered
A score, not a yes or no
A 0 to 10 score let the team start with the best prospects and go down the list, instead of drowning in thousands of maybes.
A person still hits send
The AI writes the message, and a person reviews it and sends it with one click. Fast, but nobody gets a message no one looked at.
The same data answered a bigger question
Once every new company in Brazil was in one place, I could see which sectors moved the most money abroad. That analysis is how I convinced our CEO to focus on developers, who became more than half of Ruvo's customers.
Result
About 30,000 companies went through it, mostly tech companies and solo developers, which was exactly our target. The team used it to work the best leads first, and some became customers.
Built with Claude Code, Python and MCP.