Why AI-native companies need more people, not fewer

The claim

I spent ten years at a large retail business, moving from UX into digital transformation and building AI workflows, and more recently built my own live AI platform. From both sides, the usual story about AI and jobs looks wrong to me. That story says companies adopt AI to cut headcount and keep output steady. The companies I'd worry about aren't the ones with lots of people. They're the ones using AI to shrink the organisation they already have, instead of building a new one.

Three companies, one sector

Imagine three insurance companies with the same basic business model. Company A and Company B each have over a thousand employees and a structure that predates AI by decades. Company C is an AI-native startup of about ten people, built around AI from day one.

All three adopt AI. What separates them is what they do with their people.

Where Company A goes wrong

Company A adopts AI and cuts staff to protect its margins. The structure stays the same: workflows, approval chains and ways of serving customers designed for a thousand people, now run by fewer. Cutting staff shrinks the wage bill, not the structure, so Company A still needs a large company's revenue to carry a large company's shape.

It has also cut the wrong thing. In a company that old, most of what makes the work right was never written down: the exceptions, the history with each customer, the judgement calls. That context debt lives in people. Cut the people and the knowledge walks out with them, just as the AI that was meant to replace them needs it most. A model can't ask the colleague who left.

What Company C actually proves

Company C isn't ten people doing the work of a thousand by working harder. It's ten people inside a structure designed for AI from the start: different workflows, lower overheads, a different relationship between headcount and output.

Because it was built around AI from day one, its decisions, rules and exceptions had to be written down in a form its tools could use as the company was built. There's no decade of undocumented habit to unpick. That, not the headcount, is what makes the small team work.

The wrong lesson from Company C is "you need fewer people to win". The right one is about structure, and that structure is available to a thousand-person company willing to rebuild around it.

Company B's move

Company B adopts AI too, but instead of cutting staff to chase Company C's efficiency, it does the harder thing: it trains its people and restructures around the same AI-native principles. And it puts those people to work on the debt, getting what they know out of their heads and into a form AI can use, then keeping it current.

That's slower and harder than a layoff. It's also the move that compounds. Company B ends up with what Company C has, a structure built for AI, plus something Company C can't buy: decades of hard-won knowledge about its customers, written down. Its people become far more productive than they were, not less necessary.

What this means for Company A

Company A cut its way towards Company C's size and missed that Company C's size was never the point. Company C's structure was. Now Company A is competing with a company that kept its knowledge and rebuilt around AI, with fewer people who know how the business really works, and AI tools working from whatever was left behind in writing.

The restructuring question

None of this means AI-native startups lose. It means companies like Company A face a different choice from the one they think they're making. It isn't "adopt AI or don't". It's restructure or shrink. Training and restructuring an existing workforce around AI-native workflows is a harder, slower project than a layoff, and it's the one that compounds. The companies that treat AI as a reason to need fewer people are solving the wrong problem. The ones that treat it as a reason to need more people, trained to work in a new structure and to keep its context current, are the ones about to outcompete both the giant that downsized and the startup that started small.