AI will not kill SaaS

I chose this title because the real title is a bit complex: AI is not the end of the commercial society built on the division of labour

Yes, you can create software with AI. Yes, that is work done by some people today and those people will change the way they work.

But claiming that AI will "kill SaaS" is, with all due respect, complete bs. And I think the reason people believe it reveals a fundamental confusion — between the development of tools and the division of labour.

Start with a pin

In 1776, Adam Smith opened The Wealth of Nations with a visit to a pin factory. One worker alone, he observed, could perhaps make one pin a day. Ten workers, each specialised in one step of the process, could make forty-eight thousand.

Smith's insight was not about pins. It was about the structure of productive society. When people stop doing everything themselves and start specialising — when they divide labour — the output of the whole becomes vastly larger than the sum of its parts. That principle did not just describe the economy of 1776. It is the economy. Every commercial society we have built since then is founded on it.

Smith could not have predicted the software engineer, the UX designer, the data scientist, or the prompt engineer. But he would have recognised them immediately as new forms of exactly what he described.

Tools are not labour

Here is where the "AI will kill SaaS" argument goes wrong.

SaaS — software delivered as a service, built by specialised teams, sold to organisations who use it to do their work better — is a tool. Office365 is a tool as Google Mail is a tool and so on.

AI is also a tool.

Saying AI will kill SaaS is like saying the steam engine killed the division of labour. It did not. It changed what people specialised in. It created entirely new categories of work that nobody could have named before the steam engine existed. The division of labour did not disappear — it reorganised itself around the new tool.

The confusion comes from mixing up two different things: what tools we use, and how we organise human effort. AI changes the first. It does not dissolve the second.

What we cannot yet see

The honest answer to "what does AI do to the division of labour" is: we do not know yet, and that is exactly the point.

Smith's pin factory workers did not know they were inventing industrial capitalism. The first programmers writing COBOL did not know they were creating an ecosystem that would one day employ millions of people in roles nobody had a name for yet. The people building the first SaaS products in the early 2000s did not know they were creating a job category called "customer success manager."

Every major tool shift creates new specialisations we cannot see from where we are standing. AI will not be the exception.

What I see already: people who specialise in structuring the right questions for AI systems. People who specialise in validating and calibrating AI output — because, as I wrote elsewhere, trusting AI blindly is a habit we need to unlearn. People who specialise in the orchestration of AI agents across complex workflows. None of these roles had a name five years ago.

Why "AI kills SaaS" is the wrong question

The companies building SaaS are not in the business of writing code. They are in the business of understanding a domain deeply enough to encode it into a product that other people can use without having to understand that domain themselves. That is a form of specialisation. That is division of labour.

AI might change how those companies write their code. It might compress their engineering teams. It might shift where the value sits inside the product. But the underlying logic — that specialised knowledge, packaged and delivered as a service, creates value for people who do not have that specialisation — does not go away because the implementation tool changed.

If anything, AI will raise the bar for what counts as genuine domain knowledge worth packaging. The commodity parts of software get cheaper. The deep understanding of a domain, encoded into a product that solves a real problem, becomes more valuable, not less.

The pattern always repeats

We have been here before. The spreadsheet was going to kill accountants. The internet was going to kill retail. Automated checkout was going to kill supermarket jobs. Each time, the tool changed the shape of the work. The work did not vanish.

What the tool always kills is the part of the job that should not have been a job in the first place — the repetitive, mechanical, low-judgment part that was only human because we had nothing better. And each time, that creates room for new specialisations that we could not have imagined before the tool existed.

AI is not the end of the commercial society built on the division of labour.

It is the next reorganisation of it — and we are only at the very beginning of finding out what shape it will take.