SaaS is not dead in the age of AI

The easy version of software is getting cheaper. That does not make software businesses disappear. It changes what customers are willing to pay for.

By Michael Frankland · 31 July 2026 · 8 minute read

Every few weeks there is a new version of the same argument: AI has made software so easy to build that SaaS is finished. Anyone can generate a dashboard, a landing page, or a thin wrapper around a model. Why would a customer pay for another app?

There is a useful warning inside that argument. A generic feature is easier to copy than it used to be. A product with no distribution, no point of view, and no reason to be trusted is in a difficult position.

But “SaaS is dead” is the wrong conclusion. Customers still have recurring problems. They still need records, permissions, workflows, integrations, support, and an answer when something goes wrong. The opportunity has moved away from merely shipping a feature and towards owning a small, important outcome.

AI lowers the cost of building, not the cost of caring

AI is excellent at producing a first version. It can generate components, write a query, explain an error, and turn a rough idea into something a person can click. That is genuinely powerful.

It does not remove the less glamorous work around a product:

• understanding the edge cases in a real workflow;
• keeping data private and recoverable;
• making an outcome repeatable instead of impressive once;
• supporting the customer when the generated answer is wrong;
• earning enough trust that a team will use the product every week.

As building gets faster, these details become more visible. A small publisher can compete here because focus is an advantage. You do not need to serve every workflow. You need to understand one of them unusually well.

Five places where small SaaS can still win

Distribution is now part of the product

The strongest AI-era products are not necessarily the ones with the most features. They are the ones that make their value easy to discover, easy to demonstrate, and easy to share.

That can be built into the product itself:

• a useful export that carries a small attribution link;
• a before-and-after report a customer can show to a colleague;
• a public template that solves one part of the problem before signup;
• an API or automation that lets the product travel into an existing workflow;
• a clear result that gives the user a natural reason to mention the tool.

This is not a trick. It is a product decision: make the successful outcome legible outside the product. A small publisher cannot assume a large ad budget will do that work forever.

What publishers should stop building

AI makes it tempting to produce another general assistant, another broad productivity workspace, or another collection of interchangeable automations. Those can be useful, but they are hard to position and even harder to remember.

I would be cautious about products where:

• the customer cannot describe the job in one sentence;
• the first value depends on importing a large amount of data;
• the only advantage is a model that a bigger company can switch on;
• there is no obvious place to find the first 20 users;
• the product has no opinion about what a good result looks like.

The issue is not that these products are impossible. It is that they make the publisher fight on the most crowded part of the market.

A practical test for a new app

Before writing a large roadmap, I would run this five-question test:

1. Who has the problem this week, not someday?
2. What existing tool are they tolerating?
3. What evidence proves the problem happened?
4. What is the smallest result they would pay to receive?
5. Where can I meet ten of these people without buying an audience?

If the answers are vague, more code will not make the idea clearer. If the answers are specific, AI can make the first version dramatically faster to ship.

How this shapes our small portfolio

At Yuzool, this is why we keep returning to narrow apps rather than one giant “AI platform.” Rank for Mac turns Search Console noise into a short list of pages worth improving. Revenue Bear brings App Store and Paddle-style sales signals into a calmer private feed. Dispatch Kun gives an iPhone view to founders who want their App Store activity close at hand.

None of those ideas is defensible because the code is impossible to copy. They become defensible when the workflow is clear, the result is trustworthy, the product is pleasant to use, and the distribution loop keeps producing evidence.

The market is noisier, not empty

There will be more software. Many of the new products will be forgettable. That is not a reason to stop publishing. It is a reason to be more precise about the person, the problem, and the proof.

SaaS is not dead. The lazy version of SaaS is under pressure. The opportunity for small publishers is to build something specific enough to matter, honest enough to trust, and distributed enough to be found.