Your agent can only see 45% of your company. That's the average.

New survey of 300 data and technology executives, via MIT Technology Review, puts a number on something most of us have felt but never measured: the average organization gives its AI agents access to just 45% of company data. The furthest-behind teams cap it under 30%. The leaders clear 70%. Here's the part that should change how you sequence a rollout: trust tracks access almost exactly. Teams over that 70% line report full confidence in what their agents decide. Everyone else is closer to a coin flip. Not "somewhat confident." A coin flip. Every company in the survey plans to run agentic AI within two years. Almost none of them have closed the data-access gap yet. That's the actual governance problem hiding behind every "why doesn't the agent know X" complaint I've fielded from a client. It's rarely the model. It's that the agent was never given a real door into the system where X lives. If you're in the "where do I start" phase, this is the argument for doing the boring data-access audit before you touch a prompt. If you're already live, it's the argument for going back and widening what the agent can actually see, not for buying a bigger model. Report: https://www.technologyreview.com/2026/08/12/1141032/scaling-ai-agents-with-trustworthy-data/ Comment your honest guess: what percentage of your company's data can your agent actually reach right now? Ballpark is fine, nobody's grading you.