I’ve spent most of my career investing in enterprise software. One thing that has become clear as AI has taken off is that most companies are going to need to capture far more of their own data and have far better control over it than they have today. Not just the structured data sitting in their existing applications — that will turn out to be incomplete. The more important data is the unstructured record of how the organization actually works: the emails, messages, calls, and meetings that show who is talking to whom, what commitments are being made, and where things are quietly breaking down.

That data is hard to use. Connecting systems is not enough. Companies will need a structured, relational map of their own organization — entities, relationships, commitments, and patterns — before even tailored AI solutions can do much that is reliable. Governments and the largest enterprises are buying this kind of capability from the likes of Palantir today. Almost no one else needs that degree of customization (arguably not even Palantir’s customers) and much less are they able to afford prices that often run into eight figures.

PanOps is trying to make a version of that capability available more broadly. Their product, Signal™, connects to a company’s existing communication tools and builds a knowledge map of the organization. A user can then ask questions in plain language and get answers grounded in the company’s own data. Each customer runs in its own dedicated, isolated environment on the PanOps platform; nothing is aggregated, used for model training, or sent to third-party model APIs (which is much more risky than many realize). The goal is to deliver something that is genuinely valuable and easy to implement and use without requiring the customer to stand up a technical team or change how they work.

I’ve been working with the team for several months. What has stood out is how carefully they have thought about the practical constraints. That combination of ambition and restraint is rarer than it should be.

I expect that some form of this capability will become close to imperative for companies that want to get real value from AI. At the same time, every major technology shift produces a group of organizations that understand the direction of travel and still move too slowly. The cost of being late is rarely obvious in the moment; it shows up later as a gap that is hard to close. I don’t think this time will be different.

I’m investing in and also joining PanOps as Executive Chair. In part, to help companies understand sooner how careful they need to be about putting their private information into large general purpose AI models. In larger part, it is to help those same companies realize that they need to do more and get going on this data work sooner.

panops.ai