AI Email Agent for Insurance Agencies
The Problem
Insurance agencies were manually qualifying and following up with incoming leads - slow, inconsistent, and entirely dependent on staff availability. Response times varied. Leads fell through.
My Role
Owned PM and QA for a live AI email agent handling lead qualification and drip sequencing. Wrote the prompt instructions, persona definitions, and output formatting constraints that determine how the agent behaves. Designed a GitHub-integrated QA pipeline that automatically evaluates agent outputs and generates structured feedback reports - creating a repeatable improvement loop without manual review overhead.
How It Works
The agent receives incoming leads, qualifies them through a structured conversation sequence, and routes warm prospects to a human closer. The QA pipeline runs automatically against every batch of outputs, flags deviations, and produces a feedback summary. Engineering handles the infrastructure - my job is making sure the agent actually does what the client needs it to do.