What makes AI-native firms structurally different?

AI-native firms are leaner, flatter and more senior than comparable non-AI startups. They run roughly 25% leaner, employ about half as many people three years after founding, have fewer management layers and a larger share of engineers, while raising similar funding and reaching comparable valuations.

What is the difference between the process channel and the product channel for AI?

The process channel is AI used as an internal tool, with workers using large language models to move faster. The product channel is AI built into what the company sells, so the software itself does work that used to require people. The study found that embedding AI into the product, not internal tool use, is what predicts smaller and flatter firms.

Do AI-native firms employ fewer people?

Yes. AI-native firms run around 25% leaner than comparable non-AI startups and employ about half as many people three years after founding. In service businesses that historically scaled by hiring, such as tutoring and exam prep, they run at roughly 30% of the headcount of non-AI peers.

Which businesses are most affected by AI-native structures?

The effect is largest in service businesses that historically scaled by hiring, such as therapy, tutoring and exam prep. One AI exam-prep platform employs seven people against 912 at a non-AI competitor delivering courses through human tutors.

Confidence isn’t a strategy: what two 2026 AI surveys show

24th July 2026 | AI analysis

Two surveys came out this month: one on UK and Irish mid-market firms, the other on the US and Canadian middle market. Different sides of the Atlantic. The numbers come out almost the same.

The UK numbers, from Klarus: nearly three-quarters of mid-market firms (73%) have partially or fully deployed AI, and 91% say they’re confident in their own internal expertise across the board. Yet only 10% have successfully scaled all their initiatives beyond the pilot stage, leaving 90% with at least some projects still stuck in what Klarus calls pilot purgatory. Ask people why projects stall and the top two answers, tied at 48% each, are lack of AI expertise and concerns over governance, ethics, security, and privacy. The same areas they had just said they felt confident about.

RSM’s numbers, covering the US and Canada: 97% of firms say they’re satisfied with what AI is delivering. But only 36% have it fully embedded across core processes. In RSM’s own words, confidence has outpaced capability.

Eighty-four per cent of firms expect to increase AI spend next year. But among them, the top budget lines getting cut to pay for it are analytics (43%), cybersecurity (41%), and outside advisory (40%), areas that map closely to the expertise, governance, and data quality gaps Klarus found were keeping UK and Irish projects stuck at pilot stage.

Firms on both sides of the Atlantic are confident, satisfied, and spending, but many are underestimating or deferring the unglamorous work on expertise, governance, and data quality that turns isolated wins into scaled outcomes. Both reports land on the same point in different words: what’s holding firms back isn’t access to AI, but the organisation’s ability to operate differently enough to use it at scale.