Founder Fundamentals Extra Shot — Scaling and AI

I usually write about things I have seen play out in practice and that I'm confident I'm right about. But as AI is still unfolding, this one is more thinking than observation. Let me know where you think I'm wrong.

Scale is the lowest common denominator of business success. Scaling means generating more external output — revenue, customers served — from the same input: cost and people. Direction, business models, sales, branding, execution — these all matter enormously, but they vary too much across businesses and markets to generalise. Scaling is universal. Not success-critical for every company, but success-relevant for most (except for businesses that rely on highly individual or bespoke output).

There are only two ways an organisation can scale. (1) Individually: by becoming more efficient per operational unit, whether person or AI. (2) Collaboratively: when two or more operational units become more efficient at working together.

What about growth — adding new units to achieve a better output/input ratio? It's no different from the above. It's individual scale (1) that usually carries a collaboration tax (2). Instead of improving an existing unit, you add a more performant one to raise the ratio across all units. Different method, same principle.

How does AI impact individual and collaborative scaling?

For individual scale, there is currently no better answer than AI. It increases throughput and output quality more quickly and more profoundly than any training program could. What it does not yet do — at least not the way it's used today — is expand the area of responsibility of a person or develop an individual contributor into a leader. That's where leadership comes in. For now. AI will eventually do a great job at training and upskilling, including monitoring real-world application and correcting as necessary.

For collaborative scale, the answer seems similar — but for the wrong reasons. AI allows businesses to operate with fewer people, and it seems to reduce the impulse to collaborate in person. This reduces overhead and increases efficiency. But technically, this isn't collaborative scaling. It's individual scaling. Collaboration isn't improved; it's just reduced in volume. Where AI does improve collaboration is by making digital communication better — clearer in content, better in tone. The effect is limited today, offset by an increase in volume, but it will improve. It will improve again when communication turns agentic.

But collaborative efficiency — doing things the right way — is only part of the picture. We also need to look at effectiveness: doing the right things. When people collaborate, they continuously process information to form judgement. In complex environments, where answers aren't knowable in advance, this becomes highly relevant. Collaboration stress-tests assumptions against people with different vantage points, incentives, and information — people who have accountability and something genuinely at stake. It produces something beyond efficiency: distributed sense-making under uncertainty. People who always reach for the AI answer skip the argument that would have caught the flaw.

This effect can't easily be quantified. A delta in efficiency can be measured — in units, revenue, NPS. A delta in effectiveness cannot, because the counterfactual is unobservable. You never see the road not taken.

I can't prove it's worth keeping people around who argue. But here's what I think: to scale well, make individuals as efficient as possible — but protect the productive friction in collaboration.