Enterprise agents hit the context wall
Aaron Levie’s stronger enterprise AI point today is that agent adoption runs into a context problem before it runs into a model problem. Coding agents can lean on codebases and technical users; knowledge-work agents need permissions, fragmented legacy systems, process history, decisions, and tribal knowledge turned into usable digital context.
That makes the topic globally relevant because it shifts the applied-AI bottleneck from “which model?” to “which organization can make its own knowledge legible and secure enough for agents?” It also explains why field deployment engineers, system integrators, and vertical AI companies are becoming part of the core AI buildout.
This is effectively the #1 problem for AI agents in the enterprise.