Actify

Somebody asks their assistant which running shoes to buy, and the assistant answers. That answer is a purchase decision. It carries the commercial weight a page of search results used to carry, and nobody gets paid for it.

That gap was the idea behind Actify. The layer between a person and a product has moved. The money hasn’t moved with it yet.

How it worked

An AI service that is already answering a user sends the question over to Actify on the side. If there is a purchase intent in it, Actify sends back a handful of matching products with prices and links, and the service can put them into its answer. If someone buys, the service earns a share.

Two things mattered beyond the matching itself.

If Actify was slow or unreachable, the answer went out without it. An advertising layer that can make the underlying product worse gets removed within a week, so it had to be impossible for it to do that.

And integrating it had to be one call and nothing else. The operators who would benefit most from the revenue are small, and they are the least likely to start an integration project to get it.

Why I stopped

I talked to several operators of AI services about it. It never went into production.

Part of it was my own focus: I wasn’t primarily trying to make money, and this only works if you are. The larger part is that the business model is genuinely uncertain at the moment. The assistants that will one day sit in front of most purchases aren’t sitting in front of enough of them yet, and asking an operator to bet a product decision on that is asking a lot.

If you’re working on something in this direction, I’d be glad to hear about it. Write me.