Cold Outreach OS

Marketing agencies that run cold email for a bunch of clients at once usually end up with a mess — a spreadsheet per client, a scraper someone cobbled together, no single place to see what's actually going out the door. I built the single place. It's a dashboard where an agency finds the right people to email, checks that those people are worth emailing, writes the emails, and sends them, for as many clients as they're running, from one login.

The interesting part isn't the dashboard — it's that “find leads” means three different things depending on what the client already knows, so I built three separate paths instead of forcing one search box to do all three jobs badly. Sometimes the client already knows the company and just needs the right contact and some intel on them. Sometimes they only know their ideal customer profile and need the system to find matching companies from nothing. And sometimes the smartest move is to only chase companies that just did something specific — raised money, made a key hire — because timing beats targeting, and searching broadly first just wastes API calls on companies where nothing's actually happening.

Underneath, a background worker walks every lead through the pipeline — find, check, verify, personalize, send — and every step remembers whether it already ran, so a retry after a hiccup never double-charges an API or double-emails someone. That sounds small. It's the difference between something that works in a demo and something you can leave running unattended for a paying client, which is what this actually is.

AI decides who qualifies and what the email should say. It doesn't touch the parts where a wrong guess costs real money or wrecks a client's sender reputation — those stay deterministic, on purpose, because that's exactly the kind of call an engineer has to make and a model shouldn't.

Not everything AI touches earns its place, either. An early version used a model to shortlist companies by inferred intent. It didn't work reliably, so I cut it rather than keep patching something that wasn't going to get there.

Live today, running real campaigns for real clients.