A recent investor note argued that software is going through a reset. In short, multiples are compressing, AI is lowering the cost of creation, and the assumptions that supported the last generation of SaaS are being questioned.
I think that’s directionally right, but it misses something deeper. Demand for software isn’t declining. The value of software abstractions is. AI doesn’t kill SaaS. It kills a specific, popular variant of it: the unicorn-shaped SaaS product that tries to serve everyone through the same workflow.
When Abstraction Was the Product
For the last two decades, abstraction was the business model.
Software was expensive to build and cheap to distribute, so the winning strategy was to create one generalized product and spread it across as many customers as possible. Differences were absorbed through configuration, integrations, plugins, permissions, and process.
That only works if customers are similar enough. Most aren’t. Many workflows are only superficially alike. What looked like flexibility was often a way to manage the mismatch. Users adapted themselves to the platform while waiting for the one feature that would finally make it fit.
The product became a compromise. AI changes the economics of that compromise. Software doesn’t suddenly become perfect. It becomes cheaper to create and modify. A small team can sit with a customer, understand how work actually gets done, and generate most of the implementation quickly.
The bottleneck then shifts from coding to judgment.
- Which nouns and verbs matter makes sense for this problem?
- What should always happen? What can never happen?
- What technologies suit this problem space?
And so on.
Generic Becomes a Choice
When implementation becomes cheap, the cost of the wrong abstraction becomes impossible to ignore.
Instead of asking, “Can we adapt our workflow to this tool?” we start asking, “Why are we adapting our workflow at all?” Generic platforms stop being the obvious, practical answer. The old equilibrium favored SaaS because customization was expensive. The new equilibrium makes specificity affordable. Value shifts toward software that encodes domain knowledge directly into the product. That knowledge compounds over time and becomes the product itself.
The best software will increasingly feel less like a platform and more like an extension of a particular way of working.
Business Models Change
Traditional SaaS grows horizontally: more seats, more modules, more customers. Specialized software grows vertically. The value comes from maintaining fit as a domain evolves. Revenue becomes less about seat expansion and more about adaptation, reliability, and accumulated understanding.
The relationship matters as much as the product. That becomes even more important in a world increasingly mediated by agents acting on behalf of humans and other agents.
The Reset
As software becomes cheaper, faster, and more abundant, what becomes scarce is modeling: understanding a domain deeply enough to encode its constraints, risks, and accumulated wisdom into software.
The last generation of SaaS was built on the assumption that many companies could operate through the same model. The next generation will be built on the opposite assumption: that the differences matter.
In an era where code is abundant, the competitive advantage is no longer implementation.
It’s understanding the problem well enough to know what the code should say.