Why Will AI Workflow Software Still Be Worth Trillions?

Aaron Levie says applying intelligence to real work will still be a huge software market, because cloud infrastructure and the software built on it each created trillions of dollars of value, and the same split is coming for models. A model can be superintelligence and the workflow still has to connect other data systems, keep a human in the loop, sit idle through delays, change the business process, and deal with legacy systems that have not been modernized. Five or ten blockers of that kind are much more operational than pure model intelligence, and a classic research organization is the last group that wants to attack every one of them.

AWS, GCP, and Azure created trillions of dollars of market cap in infrastructure, and software that exists only because of that infrastructure created trillions more. Ten years ago, early GCP or AWS work in the data space would not have predicted Snowflake or Databricks. The infrastructure already seemed to cover it, so another $10 billion of revenue for products that only make data easier to work with would have looked unnecessary.

The models will be insanely valuable. Bringing them into real workflows in banking, life sciences, healthcare, and government is going to be a lot of software. For the next two to five years, model providers face how far to move up the stack and compete at that layer, or leave the space open on purpose or by accident. They want to sit closer to the customer, and they want an ecosystem so people trust them.