CentreAI Lead Sourcing & Outreach Pipeline
An automated multi-platform lead discovery pipeline: source, qualify, draft outreach with Claude, then route to a human for approval before anything is sent.
- Role
- Built and iterated the pipeline end-to-end
- Timeline
- 2025 – Present
- Outcome
- 184 qualified leads sourced and drafted for outreach
The problem
Growing CentreAI meant finding language centre owners to talk to — but manual lead sourcing doesn't scale on 5-8 hours a week outside a full-time job, and generic outreach to a business's general inbox rarely reaches an actual decision maker.
What I built
I extended the open-source Agent Reach capability layer, via Hermes and Claude, into an automated lead-sourcing and outreach-drafting pipeline for multi-platform lead discovery. The pipeline:
- Sources and qualifies leads across platforms
- Has Claude draft a tailored outreach message per lead
- Routes every draft to a review dashboard for human approval before it's ever sent
Nothing goes out without a human reading it first — the AI drafts, a person decides.
The real diagnostic work
Early on, reply rates were low. The obvious assumption is "the message needs work" — but I diagnosed it as a targeting problem, not a messaging problem: outreach was landing in generic business inboxes instead of reaching a named owner. I rebuilt the qualification step to require owner identification before a lead counted as qualified, which is a real discovery → measure → iterate cycle, not a one-off fix.
Outcome
184 qualified leads and 110 drafted outreach messages generated to date, with a qualification bar that now filters for reachability, not just relevance.