Many AI agent problems are really just information retrieval problems. If the agent has a better way to find and comb through data performantly, you will get far better results. Compute is fungible so you can use it during indexing and processing or you can use later to crank through results better. You just decide based on performance and cost tradeoffs. But the cool thing about AI agents is that you always have “more compute” as a variable in making them perform better. Most forms of knowledge work have never really had this characteristic before.