- Robin van Veen - AI Automation
- Posts
- 435 warm prospects were already in my network
435 warm prospects were already in my network
I sorted 25,033 LinkedIn connections in 71 seconds for less than a dollar. You can run the same tool.
Last week I ran all 25,033 of my LinkedIn connections through an ICP check. It took 71 seconds and cost less than a dollar.
435 came back warm. Those are decision makers at the kind of company I sell to, and they were in my network the whole time.
By hand, at 10 seconds a profile, that's almost 70 hours. So I never did it.
I'm testing Jev by TypeSafe. It answers one question per connection: does this person fit my ICP? It only sees job title, company, headline and whether someone engaged with my posts, never their name.
My criteria: owner, CEO, MD, COO or head of operations. Company in or near the Netherlands, likely 50+ employees. Engaging with my posts is a plus.
The result:
435 warm
1,982 maybe warm
5,856 cold
16,756 not relevant
Two thirds of my network isn't my buyer. Good to know, and a bit painful.
My first run found 2 warm people out of 25,000. That was on me. I had made engaging with my posts a requirement, and only 205 of my connections had done that in the last 90 days. For the second run I made engagement a plus and stopped counting unknown company size as a no.
If you try this, start with your newest 100 connections. Read what comes back, fix your criteria, and then run everything.
A lot of people asked for the tool on LinkedIn, so I put it on GitHub: github.com/Arcgent/linkedin-icp-sorter
You need:
Your LinkedIn connections export (Settings > Data privacy > Get a copy of your data > Connections)
An OpenRouter API key. A full run of 25,000 connections costs about $1.
Node 22 or newer on your computer
Everything runs on your own computer, so your connections, names and results stay there. You do need a terminal. If that's not your thing, forward this email to someone on your team who's comfortable with one. The README explains each step.
This is a side project I built to test Jev. We don't use it live. I'm sharing it as is, so check the output before you reach out to anyone.
Robin
PS: At Arcgent we make companies AI-native one process at a time. We map and score your processes, plan the order, then build and run them. If you want to know where your company would start, book a 30-minute call.