There's a version of the AI narrative that goes like this: move fast, ship early, and the technology will do the rest. While AI offers incredible compounding advantages, the foundations of technology and product remain as important as ever.
At HumanX in San Francisco, Nathalie Criou, Vapi's VP of Product Management, joined product leaders from Salesforce, Databricks, and TinyFish to dig into what actually creates lasting competitive advantage in an AI-native world. The conversation kept returning to the same place: AI is raising the stakes on the fundamentals, not replacing them. As part of this panel, Nathalie emphasized that, for Vapi, a company that has processed millions of real-world voice interactions and built its platform on what that data teaches it, it's not just a theoretical debate but a lens through which we build.
AI doesn't replace domain expertise; it exposes who has it
AI is expanding what skilled people can do, compressing mechanical work and freeing practitioners for higher-leverage decisions. But the judgment to know what to build, when to trust the output, and where the model falls short still comes from hard-won experience. Deep domain knowledge remains a genuine moat, and the gap between people who have it and people who don't is becoming more visible, not less.
The same principle applies to the systems themselves. At Vapi, we've seen this firsthand: a platform built on years of real-world voice interactions doesn't just get better at understanding speech. It develops institutional memory. It learns when to speed up, when to slow down, and when to stay silent. Every conversation handled, every edge case navigated, every correction made compounds into something that can't be replicated in a single training run. That accumulated intelligence is its own moat, built from thousands of real deployments, not just from a model.
The framing we keep coming back to: augmentation and enhancing differentiated people and technology. AI makes great people and great systems greater still. But it doesn't manufacture expertise that wasn't there to begin with.
The talent bar is higher, not lower
If AI is absorbing more of the execution, what should companies actually be hiring for? Nathalie's view, shared by the other panelists, is that four qualities matter more than ever:
One of these qualities is intellectual curiosity. Since AI is trained on the past, uncovering a new market, reframing a problem, or defining a new product category still requires human drive to figure out what hasn't been figured out yet.
Another is comfort with ambiguity, ownership, and being able to blaze your own path. The best people step up when things are still undefined, rather than waiting for clarity that may never come.
The panel also discussed how genuine customer obsession drives the kind of unreasonable effort AI can't replicate. AI still doesn't care.
And lastly, a driver 's-seat mentality. The people winning right now understand how to use AI to amplify themselves, not be carried along by it.
The implication is that the execution gap between strong and weak performers is narrowing. The judgment, creativity, and ideas gap is widening.
The moats that survive AI, and the ones that don't
Here's the most important thing to understand about AI and competitive advantage: it doesn't dissolve moats; it helps identify which will be the most enduring. The moats that were always somewhat shallow, such as being first to market, holding a rare technical skill, are eroding fast because AI makes them easier to replicate. The moats that were always genuinely deep are compounding faster than ever, because AI gives the leaders more surface area to extend their advantage.
The moats that hold up are the ones tied to trust, switching costs, and data that can only be generated by actually doing the work:
Systems of record endure because every integration in an enterprise reads from and writes back to them. Switching costs are enormous and grow with every passing year.
Distribution and customer relationships matter because trust built through previous platform shifts, cloud, mobile, and social is hard to manufacture. Companies default to partners with a proven track record.
Community and mindshare are nearly impossible to replicate from scratch. An ecosystem that's emotionally and professionally invested in a platform doesn't migrate easily.
Unique interaction data may be the most durable moat of all, and the one most specific to AI-native companies. This is not created through training data but through the signal generated by real customer interactions at scale, the kind that teaches a voice when to speed up, slow down, or stay silent based on who it's talking to. As Nathalie put it: "A lot of this can only be learned by doing, and doing it takes time, which is really all that a moat can give you."
The Bottom Line
AI is a powerful amplifier of great people, great products, and great businesses. But it's not a shortcut around the fundamentals; it's an amplifier of the foundations that a product has created. The companies building durable products in this era are the ones doubling down on domain expertise, hiring for judgment and empathy, and investing in moats that compound over time.
At Vapi, that's not a strategic position. It's what we've learned from building in the field.
