Case study · AI automation
AI agent for lead generation.
Autonomous research-and-outreach loop that replaced a recurring Friday afternoon.
The brief
What they came in with.
An internal project. We built the system we wished our clients already had. The brief: compress a manual weekly research-and-outreach workflow into a Monday morning review.
No demos that never ship. Just a working loop with real integrations and measurable reclaimed time.
The approach
How we think about it.
Every project starts with the business problem, not the solution. Before touching any tool or writing a line of code, we spend time with the people who own the outcome. We map what's broken, what's actually worth fixing, and what success looks like in practice, not in a brief. Only then do we build.
We mapped the workflow on paper first. Six steps, three decision points, five tabs always open on a Friday. Once we had it written down, the automation was obvious.
Boring stack on purpose: n8n for orchestration, OpenAI for synthesis, Supabase for the prospect ledger.
What we shipped
The work, in plain language.
- 01n8n flow runs the schedule, handles retries, and keeps credentials in one place
- 02OpenAI Enterprise endpoint with no-training clause drafts first-touch messages in the client's voice
- 03Supabase table stores the prospect ledger so the loop doesn't repeat itself
- 04Defensive fallbacks on every field; alert fires when a fallback is hit
- 05One 30-minute maintenance check on the first Monday of each month
The Friday isn't a bad day anymore. That's the whole pitch.