50-70% of companies on a typical outreach list don't actually fit your ICP. I see this all the time, both at our agency (20+ salespeople) and from conversations with other teams. Doesn't matter if you're using Apollo, Clay, or ZoomInfo - people build lists filtering by industry, company size, maybe location, then think that's job done.
But those filters only get you in the ballpark. You search for 'fintech' and half the results are payment consultancies tagged as fintech. Or a company pivoted a year ago and the database still has the old description. Stuff that technically matches your filters but is obviously wrong the second you actually look at what the company does.
Most teams handle this one of two ways: either they blast the full list and wonder why reply rates are terrible and domain reputation tanks, or someone sits there for a day clicking through websites manually.
So we built an open-source setup where an AI agent does this check for every single company on your list. It scrapes the actual website, reads what the company does right now (not what the database thinks), and writes a yes or no with reasoning.
Recent campaign starting with about 900 Apollo results - the agent went through all of them, threw out around 600. The 300 that survived were actually solid. That's typical for us: somewhere between 50-70% gets filtered out depending on how niche the ICP is.
We open-sourced the setup if anyone wants to poke around. It runs inside Claude Code with 13 skills that walk the agent through each step. You plug in your own Apollo and SmartLead keys and everything runs locally - we don't see any of your data. Not trying to sell anything, the open-source version is genuinely free with no upsell page anywhere. Would appreciate any honest takes, even brutal ones.
Curious how other teams here handle list quality. Did you build an automated pipeline, do you do it manually, or are there tools on the market?
