AI War for Enterprise

As Claude takes off as the de-facto AI-first Work OS, there are a few implications worth thinking through. Let's walk through some traditionally well understood dimensions of enterprise software and see how an AI native OS may play out along them.

What matters is compliance & utility per unit time.

There are two important observations here. First, model providers clearly have the smartest models. It's possible that there is diminishing marginal utility with new model intelligence, but that won't be the case for at least a year or so IMO. The implication is that model providers have a huge head start on utility. Second, marginal utility grows with each new data connector. 4 data connectors are more than twice as valuable as 2 (because you can cross-reference information). This means there will be deep retention, and the head start will count so much more. Third, compliance is table stakes for enterprise adoption — the AI OS that can demonstrate audit trails, data residency controls, and policy enforcement across all those integrations will win the trust of procurement teams and CISOs, which in practice means winning the deal.

Positioning is still important.

A sticker that labels your software as 'X for A' is generally understood to be much, much more effective than just selling something as 'X'. You might argue that there will be many 'Claude for industry A', but for the reasons mentioned above, Claude (or one of a few competitors) will become the de-facto OS. We'll see MCP integrations as the new competing ground. Rather than offering 'Claude for finance in X', enterprise apps will offer 'Finance tools for X as MCP'. There is lots to do here in building out effective AI integrations — data readiness, compliance controls, and offering flexibility in the MCP layer. I'll likely write another post on this.

Enterprise sales cycles are long and involve a lot of human input — but that might not matter as much.

I think the traditional cycle will be largely circumvented by a bottoms-up approach. If you see your job being made 50x more efficient (and so the chance for 50x more impact), you're not going to let bureaucratic friction get in the way. We're going to see much faster enterprise sales cycles, relatively speaking.

So what does this mean?

Whether you're thinking about leveraging AI for your own business, or a software business building for the enterprise, you need to start spending all your time on understanding the business workflows and metadata that allows you to provide specialized context to the AI OS.