Spotting User Pull in AI

In early AI, the real edge is spotting whether users are pulling a product into daily workflow before the model layer gets commoditized. That is why Tribe leans on normalized company data instead of category narratives. At this stage, the key question is not whether a startup uses a frontier model, but whether users come back, activate fast, and bend real work around it the way Gamma did after AI cut the blank page problem, or the way Calendly spread by making one painful task obviously easier.

• Tribe describes its underwriting process as building a database level view of companies, normalizing metrics across thousands of startups, and comparing them apples to apples. In AI, that matters because market maps change faster than user behavior does.
  1 sacra
• A read on product market fit is concrete. Gamma saw activation change from a weak blank canvas experience to users getting an aha in minutes, then signups jumped from hundreds a day to about 10,000 a day and later 50,000 in a day. That is pull, not hype.
  1 sacra 2 sacra
• Some of the best software wedges looked small at first. Calendly spread because every booking link showed the product in action, and Deel found fit by turning messy contractor payments and compliance into a one click payroll like workflow. In both cases, repeated usage made the market look bigger over time.
  3 sacra 4 sacra

As AI markets mature, the winners are likely to be the companies that turn model output into habit, workflow, and distribution. The next step is less about having access to a better model, and more about proving that users return often enough, or depend deeply enough, that the product keeps its place even when the underlying intelligence becomes widely available.