The Leads Weren't Dead. The Math Was.

I was traveling when a friend called. He’d found a company that wanted to point AI at the leads nobody wants to work.

This company got plenty of leads. The qualified ones went straight to the sales team. Under that sat a pile of 523 that never got qualified — people who raised their hand once, but nobody knew if they were real, and nobody had time to find out. They’d even tried to build something to work them. Couldn’t pull off the tech, so they let it die.

They didn’t decide those leads were worthless. They decided working them wasn’t worth a person’s time. Two different things — and that difference turned out to be everything.

My buddy pulled me in for the technical build. That’s the part I do. We had a week. First we drew up the “right” way — plug into their CRM, their tools, every system they had. We killed that in an hour. Too many moving parts, too easy to break, no way we’d ship it clean in time. So we did the opposite: one small thing that did the whole job by itself.

One Cloudflare worker. One connection — WhatsApp, where their leads lived. One big prompt carrying the whole sales script in a single shot. It was rough. Stateless, heavy-handed, none of the clever stuff. I built it, my buddy tested it and handled the client. Three days.

Then we pointed it at the 523.

Over three weeks, 17 of those written-off leads turned into real, qualified conversations. The bot talked to them, qualified them, and handed them to the human sales team. The team closed about $2,000 out of that pile. The whole thing cost around $200 in tokens. The bot never closed anyone and never pretended to be a person — it qualified and handed off. People did the closing.

Two lessons, pulling in opposite directions.

Dead leads aren’t dead — but it’s bigger than a motivational quote. The math changed. It never made sense to pay a salesperson to chase an unqualified lead for a week; their time cost more than the lead was worth, so the leads just sat. But a bot can work the whole pile, all day, for pennies. What made those leads “not worth it” was the cost of human attention — and that cost just dropped through the floor. A pile of dead leads isn’t dead weight. It’s money you already spent, sitting in a drawer, waiting for the cost of working it to come down. It finally did.

Simple and shipped beats fancy and late. The clean, fully-integrated version would still be half-done. The dumb version shipped in three days and made money. I learn this, forget it, and learn it again on the next project.

It’s the first one I keep coming back to. Most companies with a sales team are sitting on their own version of those 523 leads — generated, paid for, quietly written off because working them by hand never added up. I don’t think that’s true anymore.

So that’s what I’ve been chasing lately: where else does the old math quietly not work now?