AI didn’t kill SaaS. It changed what people will pay for.
“SaaS is dead” makes a good headline. What is actually happening is narrower and more useful to understand: AI is shrinking the value of some software and growing the value of the rest.
October 23, 2026 · 4 min read
Every few months someone declares that software-as-a-service is finished. The argument goes like this: AI can now write working software in an afternoon, agents can click through apps on our behalf, so why would anyone keep paying a monthly fee for a tool?
It is a fair question, and parts of it are right. But “SaaS is dead” is too blunt to be useful. What we see, building products for startups every week, is that AI is pulling software apart into two groups: products whose value was mostly the interface, and products whose value sits underneath it.
What AI really changes
Three things have shifted in the last two years, and they matter more than any single model launch:
- Building is cheaper. A small team, or a single founder, can now ship in weeks what used to take a funded team months. The cost of copying a feature has dropped close to zero.
- The interface is optional. If an assistant can read your data and take the action, the screens a product was built around matter less. People ask for the result instead of navigating to it.
- Seats stop making sense. Many tools charge per person. When one person with an AI agent does the work of three, per-seat pricing starts to shrink on its own.
Each of these is real. Together they are a problem for a specific kind of product.
The products under pressure
- Thin wrappers. Tools that put a nicer screen on top of data that lives somewhere else, or on top of someone else’s model, are the easiest to replace.
- Simple single-purpose utilities. A form builder, a basic scheduler or a one-feature dashboard can now be generated, or handled inside a larger tool’s assistant.
- Products priced on headcount alone. If the value grows with the work done but the price grows with the number of people, the two drift apart.
- Products whose main job was workflow. Moving a task from one column to the next is exactly what agents are getting good at.
The products getting stronger
The other side of the split is less talked about. Some software becomes more valuable because of AI, not less:
- Systems of record. The place where the accounts, customers, contracts or patient notes actually live. Agents need a trustworthy source to read from and write to, and moving that data is risky.
- Products with their own data. Years of transactions, usage or outcomes are hard to copy, and they make an AI feature better than a generic one.
- Regulated and high-trust work. Payments, healthcare, payroll and tax carry licences, audits and liability. A weekend clone does not come with any of those.
- Network effects. Marketplaces and tools where every new user makes the product better for the others.
- Deep workflow knowledge. Software built around how a specific industry really works, including the exceptions, beats a general assistant that knows a little about everything.
AI makes features cheap. It does not make trust, data or distribution cheap.
Pricing is where the change shows first
The clearest sign that SaaS is changing rather than ending is pricing. More products now charge for what gets done (a resolved ticket, a processed invoice, a completed task) instead of, or as well as, the number of people logged in. Usage-based and outcome-based pricing is harder to forecast for both sides, but it lines the price up with the value, which per-seat pricing increasingly does not.
What this means if you are building a product
- Own something. Data, a workflow, a licence, a community, a distribution channel. If a competitor could rebuild your product from screenshots, that is the risk to fix first.
- Build AI into the job, not onto the side. A chat box bolted onto the corner of the screen rarely changes much. An assistant that does the boring half of the task inside the flow does.
- Design for people and agents. Clean APIs, clear permissions and good audit trails matter as much as the screens, because more of your users will reach you through an assistant.
- Keep the interface for the decisions that matter. Approvals, reviews and anything with money or risk attached still need a clear, calm screen a person can trust.
- Ship and learn faster than before. If building is cheaper for you, it is cheaper for your competitors too. Speed of learning from real users is now a bigger edge than the size of the codebase.
So, is SaaS dead?
No. Software people pay for every month is not going away, but the bar has moved. Products that were mainly a screen on top of a spreadsheet are in trouble. Products that hold valuable data, carry real responsibility, or know an industry deeply are well placed, and AI makes them more useful.
The question for founders is not whether SaaS survives. It is whether your product is the kind an assistant replaces, or the kind an assistant depends on. The second kind is still a very good business to build.
