if i see one more saas is dead post

Published on 20 April 2026 - Author: David Cheal

The people declaring SaaS dead don't understand what SaaS actually is. The app you see is the tip of the iceberg — everything that matters sits underneath.

There are two types of people who say that SaaS is Dead:

Those ignorant of what SaaS is
Wilfully incompetent grifters selling AI bullshit

I am really hoping most people are in the first group. At least they can be brought up to speed. Either way, it’s going to end up with lots of people in tears.

The term "Software as a Service" is counterintuitive. The Software is far less important than the Service. SaaS is actually about the economics of scale. It's intended to take something that costs a fucktonne of money to build, maintain and deliver, then make it available at a sane monthly price.

When you use a product like Salesforce, Canva, Google Docs or Spotify, you see a thin surface view of what the service does. Many people would say something like “Google Docs is a word editor in the browser” or “Spotify is a phone app that lets me stream music.”

I recently had someone tell me Amazon Kindle delivered text files (books) and I couldn't decide if I should laugh or cry.

SaaS is not an "app"

People have the most superficial understanding of what delivers modern systems. They think it's this:

Iceberg tip poking above water inside a glass cube, labelled with the visible parts of SaaS: UI and frontend, mobile and web apps, dashboards, reports.

Then they spin up Claude Code or Codex and ask for a "CRM with charts" and declare "SaaS is dead"

If they had any understanding of software development, infrastructure or service delivery, they would realise this is ridiculous. Their AI created monstrosity is missing 90% of the things that are required to deliver even a basic service.

And no, you can't just leave all these things out because the app is just for your company, or a few departments. What you've built isn't a solution, it's a liability.

Service delivery is insanely complex and difficult

You need to understand that software requires a huge amount of supporting technology

The same iceberg viewed deeper into the cube, with the submerged bulk labelled: multi-tenancy, data architecture, security and compliance, reliability and resilience, APIs and integrations, automation and workflows, infrastructure and operations, observability, continuous delivery, billing and entitlements.

Without that stuff, what you've built is a brittle, shit solution, that will fall over the moment anyone even looks at it funny. Nobody is monitoring it, ensuring it's secure, that it can handle load or that it's compliant with the regulatory requirements of your industry.

Creating viable services, involves more moving parts than you can imagine. If you haven't done it for a living, you can't even begin to understand how difficult it is. It's about SCALE

The biggest knowledge gap is scale. That service, you use, is part of a massive, interconnected platform that often spans the entire planet. To say that something like Salesforce or Kindle is just software, is beyond ridiculous.

You think you're reading this article on a website? That a server somewhere just spat out this post? LinkedIn is built from tens of thousands of microservices distributed across the planet, that handle billions of requests per day. There are at least 4000 engineers building and maintaining this SaaS platform every day.

Cross-section of a vast iceberg whose underwater base dwarfs the small peak above, labelled: global reach, massive scale, multi-tenant efficiency, elastic infrastructure, economies of scale, compounding advantage.

SaaS takes something that costs 100's of millions to build and run, then delivers it to you for $50 bucks a month. This only works because the financial modelling allows it. At scale.

The AI-created app, you think kills SaaS, is thousands of times more expensive compared to the SaaS. Adding even basic one for one capability will drive the cost through the roof.

If you want a good example of the cost gap, just consider the risks you are blindly accepting. SaaS providers assess the risks and mitigate them. They ensure there are minimal outages, for instance. What's the cost to you or your business if your AI made CRM falls over for a day? What's the cost of a lost sale due to an outage?

Even if you are building software just for your own personal use, SaaS is still a better option. You will waste far more time trying to create the software than the cost of just buying it.

It's also a little ironic that products like Codex or Claude Code that apparently kill SaaS, are in fact SaaS...

AI is not the problem

To be clear, this is not a problem with AI coding. AI can 100% help you build a SaaS solution. It can help you build anything. But you have to understand what you are building. What actually needs to be created and delivered so you, your boss or your clients can get a viable result.

If Claude Code built a solution in a few weeks, and you think that replaces a SaaS, then you have failed. You've built a clusterfuck waiting to happen and don't even realise it yet.

In the very near future, you are going to discover that technology breaks and that lawyers litigate.