Said another way, agents are not replacing systems of record. They are raising the standards for what a good one looks like.

The more we automate, the more important it becomes that someone has done the unglamorous work of deciding what the correct answer is and where it lives.

Humans can hold nuance in their heads. If finance and sales disagree on ARR, they can talk it out and decide which number to use for the board deck. Agents cannot. They need explicit rules. They need conflict resolution baked into the data model. They need to know that “official_arr” is the metric we use for external reporting, “sales_arr” is the... See more

Said another way, agents are forcing us to separate the UX of work from the source of truth for work. The UX can now be a chat window, a natural language interface, or a specialized agent UI that looks nothing like the old enterprise applications. Under the hood, though, something still has to say “this is the canonical customer record” or “this is... See more

That combination means agents are only as good as their understanding of which system owns which truth, and what the contract is between those truths. This is the bull thesis on a company like Databricks - they become the center of gravity for AI Agents, and start to build these agents themselves.