Why Software Is Dead

Photo by Tony Pepe on Unsplash

The software in trouble is the kind we subscribe to because building our own used to be a pain. For everyday personal tools, that bargain is changing. For the systems a business runs on, I think the obituary is premature. B2B software is still alive and strong.

Pieter Levels argues for an era of personal software, calling software dead in his October 11, 2026 post. There is something useful in that provocation. There is also a big difference between replacing your personal habit tracker and replacing the system that tells a company what it can ship.

We should probably make that distinction before cancelling every subscription.

Everyday software is easier to replace

Suppose you want a small tool that compares your grocery spending from month to month. You know what you want to see. You can inspect the result. If the chart breaks, your afternoon gets slightly more annoying.

Does that need to be a product with a monthly subscription? Maybe. But if AI helps you build something that does exactly what you need, the subscription has a harder job to justify itself.

This is the appealing part of Levels's personal-software argument: a tool built for your circumstances, without waiting for a vendor to decide that your request belongs on its roadmap. It can be small because it only has to serve you.

I don't think every everyday SaaS product disappears. Plenty of people would rather pay than configure anything. A service can also have useful data or a community that you cannot reproduce by generating a new interface.

Still, selling a simple feature becomes a more awkward proposition when the customer has another way to get it. Adding an AI button does not automatically settle that problem. Everybody else can add one too.

Personal tools can fit the task

Geoffrey Litt explored this possibility in his 2023 essay on LLMs and end-user programming. He imagined one-off scripts and interfaces shaped around a task, with more people adapting software themselves. That was a speculative design argument, not evidence that subscriptions had already been displaced.

I find that more useful than assuming everybody wants to become a developer.

Someone asking where their money went probably wants an answer, not a new hobby. If an assistant can produce a useful comparison and explain what it counted, good. The person may never care whether the result came from a script or a conventional app.

But the jump from “give me an answer” to “run this part of my company” is substantial. The number of people affected changes. So does the cost of being wrong.

B2B is still alive and strong

Benedict Evans draws a useful distinction in AI tools and transformation. An improvised tool for a one-off task is different from a repeated workflow used across an organization. Once revenue or risk depends on it, maintenance and organizational requirements matter. AI can change the point at which building makes sense, but adopting a working system still involves more than generating its code.

That is his analysis, rather than proof that any particular vendor will survive. It is also a much better starting point than assuming every subscription buys the same thing.

Take a hypothetical manufacturer. One employee builds a tool to flag differences between yesterday's orders and warehouse stock. Useful. Let them.

Photo by Ashley on Unsplash

Now the sales team starts relying on that tool to promise delivery dates. Warehouse staff need to record damaged goods. Some inventory is already reserved. Who is allowed to change those records? Which system has the final number? Can somebody find out why an order was changed last Thursday?

The company has acquired a shared operational dependency. Recreating the screen is a small part of replacing it. The integration with other systems has to remain correct, and somebody has to be accountable when it fails.

A buyer may happily pay a vendor to handle that work. They are trying to manufacture things, after all. Taking on another system to maintain is not necessarily progress.

There is also actual business behind the argument. ServiceNow's Q2 2026 results, for the quarter ended June 30, reported subscription revenue of $3.877 billion, up 24.5% year over year. Salesforce's Q2 fiscal 2027 results, for the quarter ended July 31, 2026, reported subscription and support revenue of $10.8 billion, up 12%, including $440 million from Informatica. That second figure is not an organic-growth claim.

Two companies do not represent the entire market, and their results cannot settle what AI will do next. They do show substantial B2B software businesses growing while the industry is supposedly dead.

The B2B label will not save a weak product

There is a catch. A tool does not become essential because the person paying uses a company credit card.

A small report generator inside a business can face the same pressure as a personal app. If replacing it causes little disruption and nobody depends on its particular data or service, the buyer has good reason to look for a cheaper option.

The useful distinction is how deeply the software participates in the customer's work. Does it hold records other people rely on? Does it make a difficult integration dependable? Who handles the consequences when it gets something wrong?

Those questions also make the opportunity clearer for builders. Use AI to ship faster. Keep the scope small. Then find out whether the customer needs an occasional result or a service they can rely on every day. The promise you make should match the responsibility you are willing to take.

So yes, some everyday software is becoming easier to replace. B2B operational software still has a strong reason to exist. Treating both as one funeral makes for a good headline and a bad product strategy.

If a customer could recreate your interface tomorrow, what would still make replacing your business a bad idea?