The Easiest Sale in the World

A chatbot user spends twenty minutes describing their relationship with alcohol. The cravings, the relapses. At the end of that conversation, a treatment provider who specializes in exactly their situation appears. That’s the easiest sale in the world.

This match can’t happen on keywords. Keywords compress every advertiser into the same bin and extract rent from the collision. The simulation shows the cost: specialists lose -0.807 per round paying for queries they’ll never close. More than 50% of small businesses have been priced out of Google Ads entirely. Chatbot conversations are richer, longer, and more specific than search queries. Embeddings let each advertiser describe what they actually sell and match to the people who actually need it.

I’m building the ad network to make this happen, starting with health and health-adjacent chatbots.

The Second Easiest Sale

Conversational AI platforms are burning cash on inference with no ad revenue. Perplexity made $20,000 total before killing ads entirely. Vertical chatbots have deep user engagement and zero monetization infrastructure. They need revenue.

So I walk into a chatbot publisher’s office and say: you’re losing money on every conversation. I’ll pay you every time a user talks to your product. Here’s a vertical exchange with specialist advertisers. CPM beats your banner ads and pays more than inference costs because every advertiser is relevant. Zero to hero. You don’t need Google-scale liquidity. You need enough relevant bidders to clear the auction competitively, one vertical at a time.

The Third Easiest Sale

Treatment providers are paying $75 per click on “addiction treatment” keywords they share with twenty competitors. Most of those clicks don’t convert because the match is wrong. Embedding auctions match on what the provider actually does.

So I call a treatment provider and say: here’s a channel where every lead already described exactly what they need. Better unit economics than Google Ads. No commitment. Try it on 10% of your spend.

The Fourth Easiest Sale

The scoring function is published. The simulation is open-source. PubMatic already has the NVIDIA partnership for sub-millisecond inference. Anyone can reimplement this in weeks. First mover advantage is thin when the protocol is a commodity.

So Google walks into my office and says: nice proof of concept. Here’s an acqui-hire. I have to accept because there’s no moat and they have way more liquidity. If I don’t build fast enough to clear real ad traffic, Google will hit me with a job offer I can’t refuse. It’s way cheaper for them to hire one person than to acquire a company with live revenue. The offer isn’t a gag order. It’s a sedative. The protocol survives either way.

But that’s the whole point. Consumers get conversations that stay private by design. Advertisers reach the right people on their own terms instead of bidding on keywords they share with everyone else. Chatbot publishers can monetize independently and survive without selling to a larger player or bending to a single large customer. Free and open competition is the force for good.

Written with Claude Opus 4.6 via Claude Code. I directed the argument; Claude drafted prose.

Part of the Vector Space series. june@june.kim