Enterprise Content Emerges as Agentic AI Bottleneck, Box State of AI Report Says

My take: The growth of AI agents in enterprise is real and accelerating: 83% of organizations already have them in production, and the share describing themselves as advanced or leading-edge jumped from 8% to 64% in a single year. But adoption is outpacing governance, and that has concrete consequences: nearly half of companies (49%) have already experienced an AI-related data exposure incident, and only 36% have connected their agents to trusted internal content.

The problem is not the models: the models are there. The problem is the knowledge infrastructure that feeds them. An AI agent that cannot access the right company information, or that accesses it without proper controls, does not work better: it works with more risk.

If your company already has AI agents, the question is: who decides what data they can access, and who verifies that those rules are being followed?