Why the best CPO or CTO for your AI company might not be an AI native

Some observations from conversations with founders and C-suite leaders — and a few questions that keep resurfacing.

When advising founders and C-suite leaders across AI and deeptech, the same commercial tension surfaces repeatedly. The pipeline looks healthy. Early conversations go well. But enterprise customers — the ones that would genuinely move the needle — aren't committing. The assumption is usually that more sales resource will fix it. From what I've seen, that rarely gets to the root of it. The friction tends to sit somewhere upstream, and it raises some questions worth sitting with.

The enterprise AI market has a language problem. "Intelligence layer", "agentic automation", "context-aware orchestration" — these mean something precise to the people building them. It's less clear that they land with the CFO who has to sign the contract, or the COO who has to justify the decision upward. There's a related question underneath this one: how does a cautious buyer defend a seven-figure AI commitment when the model layer is moving as fast as it is? What a leading model couldn't do in January it can do by March. If a product pitch rests on today's capability, a careful buyer has reasonable grounds to pause. The vendors who seem to be cutting through are pitching something more durable — not what their AI can do, but what their architecture makes possible regardless of which model sits underneath it. Whether that framing holds up under scrutiny is a question worth asking internally.

Enterprise buyers in operationally complex industries — manufacturing, logistics, maritime, field services — tend to have long memories. Integrations that took eighteen months and delivered tools nobody used. Vendors acquired and sunset before the ROI arrived. That experience doesn't go away when a new vendor walks in with a compelling demo. It's worth asking whether your sales process acknowledges that history at all — and whether there's a way to let the product speak before a contract is on the table, particularly in industries where handing over operational data to an external platform is itself a significant ask.

This is the tension I find most interesting, and I don't think there are clean answers yet. Most AI vendors are pricing intelligence like software — seat licences, annual commitments, renewal cycles. That made sense when the value was access. When the value is operational outcomes, the model feels like a mismatch to buyers on the other side of the table, even when they can't always put their finger on why. Consumption-based pricing seems like the obvious alternative, but it creates a different problem for a CFO trying to set a budget. How the best CPOs are navigating the gap between those two models — predictability for finance, growth tied to actual value — seems like one of the more consequential product questions in enterprise AI right now.

These aren't observations about what companies should do. They're patterns that keep appearing across conversations with leadership teams at very different stages and in very different markets. What they point to is a question about the kind of CPO and CTO that can actually navigate this moment.

The instinct is often to look for leaders who are native to the AI era — and that fluency matters. But the problems above aren't new problems dressed in new technology. Enterprise caution, pricing model transitions, the gap between a compelling demo and a signed contract — these are patterns that anyone who lived through the SaaS transition, or the move to cloud, will recognise. The CPO or CTO who brings that cycle experience, and can still get hands-on with AI deeply enough to earn the room's respect, is a harder profile to find than it sounds. In my experience, it's also the one that tends to make the difference.