One of a dozen.
There are perhaps twelve omni agents left standing in the enterprise. Three belong to labs. Two to platforms, and they're captive. The rest are neutral agents that climbed from somewhere, and one of them climbed from a content store in Redwood City.
Box Agent holds the working context of 40% of the companies it serves: not just the files, but the parts nobody ever wrote down and told only to the agent. The store underneath it is now plumbing you own. Seat revenue is 8% of the business. The stock is 2.6x the 2026 price. Salesforce offers 4x that.
You did the thing the thesis says incumbents can't: killed your own pricing, stayed neutral toward everyone's data including your own, and let the app become a view. The acquirer at the end isn't a rescue. It's the incumbent that didn't do this, buying the one that did.
One path that gets here
A · Ship the agent as a per-seat add-onB · Reverse. Open the store to every agent and stop pretending Box AI is the front doorB · Use the position. Your store sees every agent's request. Build the agent on top of your own gravityA · Keep building. Route to all three labs; stay neutral on everything but the customer