Which moats still matter in an AI-native world?

Last week I had dinner with a group of software founders and investors.

Same table. Completely different energy.

The founders were electric. New use cases emerging daily. Demand they can’t keep up with. Things being built that simply weren’t possible 12 months ago.

The investors were far more cautious. The SaaS playbook that worked for decades suddenly looks less certain.

Lower barriers to entry.
Faster commoditisation.
Pricing pressure.

Several were openly stepping back from software investment — unsure where durable value actually accrues in this new world.

I understand the concern. And honestly, no one fully knows how this plays out.

But some things about great companies never change:

They solve a real problem.
And they become very hard to replace.

In an AI-native world, four advantages still hold:

  1. Proprietary data: Years of real interactions in a specific domain. Impossible to replicate.
  2. Institutional trust: Governments and enterprises don’t switch because of a slick demo. Trust is slow to earn — and sticky.
  3. Sovereignty: AI is becoming critical infrastructure. Who you are will matter as much as what you build.
  4. Distribution: When products are cheap to build, attention becomes the bottleneck. Distribution is now the scarce asset.

We’re seeing AI-native companies reach extraordinary valuations.

But I worry that many — whilst clearly solving a problem — aren’t necessarily yet hard to replace.

Yes, you have to run fast.
But what you build while running fast is what matters.
Otherwise success — and valuation — will be fleeting.

Curious how others see it: Which of these moats matters most in the AI era? And are there moats I’m missing?

Originally posted on LinkedIn.