Groundbait: an AI-driven outbound engine

The problem

Most outbound tools either blast generic messages at scale or require a salesperson to research and write every piece of outreach by hand. Neither works well: spray-and-pray damages reputation, and manual outreach doesn't scale.

What we built

Groundbait is an outbound engine that sits between the two. It researches a prospect automatically, drafts a message tailored to that specific business, then puts it in a review queue so a person reads and approves it before it goes anywhere. Nothing sends without a human sign-off.

The system covers two channels:

• Email: prospect research and personalised first-touch drafts, routed to a review queue in n8n. The reviewer edits, approves or discards each one.
• Voice: an AI phone agent built on Retell, with Claude as the reasoning layer, running a structured sales script. Calls are reviewed and approved before dialling.

The compliance-first architecture — a person in the loop at every send decision — is deliberate. It keeps the system off spam blacklists and keeps a human accountable for every message that leaves.

Why we built it

We use it to run outbound for our own products and to test the approach before offering it as a service. It is the same system we would build for a client: auditable, controllable, and designed so the team running it stays in charge of what goes out.