The context
Agentic Builders is a venture studio that starts and backs early-stage products. Each of them has to find its first acquisition channels. My question was how much of that growth work could run as software.
My role
I designed and built the pipeline myself: a multi-agent system in Python that works like an AI CMO for the studio's startups.
What I built
- Research agents that study a product's competitors and audience.
- A synthesis step that combines that research with a library of acquisition mechanics and proposes growth tactics a team wouldn't find by searching.
- AI respondent panels that give a first read on positioning, value proposition and pricing before any customer interviews.
- Support for OpenAI, Claude and Gemini models, with retrieval over a curated growth knowledge base so the agents work from real material.
What came out of it
Around 15 early-stage products ran its tactics through their acquisition work over about a year. Together they added a few thousand sign-ups.
The first pass on positioning and pricing used to take weeks of scheduling and running interviews. With the respondent panels it takes about a day, and the interviews go toward testing a sharper hypothesis.
I can't isolate the pipeline's share of those sign-ups. Teams sometimes ran a tactic as generated, sometimes adapted it, and sometimes used it as a starting point for a better idea.
The open-source version
I open-sourced a public version as Zero CAC Growth, a plugin for Claude Code and Codex. Give it a product URL and it maps the market and suggests acquisition tactics that don't depend on ad spend.