Thoughts on the SaaSpocalypse

This post focuses on the many risks that LLMs pose to SaaS businesses, while purposefully ignoring the benefits, solutions and opportunities. The post also doesn't cover doomsday scenarios where the AI becomes sentient and eliminates us (we have bigger problems than selling online software at that point), or societal/economic collapse. For those of us that have worked in SaaS over the last decade(s), we know it's an amazing business model where you build software, stick it on the server, and charge a repeated monthly or annual subscription for/based on usage - when done right, as long as your incremental new customers > churned customers it's a machine that keeps growing and building momentum. However, for many of us this machine will start coming under pressure due to the new realities caused by LLMs - and it's not just the CRUD apps with a UI over a DB or LLM wrappers at risk.

A (nightmare) scenario:

One of the best areas to be in historically for SaaS is the System of Record for a business process: CRM, Financial Accounts, HR data etc... Once the data was there, customers weren't going to leave any time soon.

Now let's picture big SaaS CRM - The contract is coming up for renewal, they've had to whack up their prices due to inflation and to prop up the under pressure share price.

The client can create some screenshots and ask their AI agent to design a system based on the application structure. You can also give the AI access to your browser so it can "learn" (take screenshots/scrape dom) all the areas of the app, what form validation is in place etc...

The client reviews the system spec, and corrects/adds any features and requirements. The client (or prospect) doesn't have to replicate your whole system - just the parts that are useful and relevant for them.

The AI then builds the whole application in a few hours-days and deploys it. What about the historic data though? they're not going to want them to lose that (your old moat)! The AI can use the APIs you provide to download that data and upload it into the new application allowing for a clean switch over. You could try to not offer APIs but customers won't like having the data locked-in.

The risks

Let's review the risks posed by LLMs and see why SaaS businesses are under pressure. Some of these are minor and some potentially existential:

Fighting for budget where the budget is being assigned to AI

AI is sucking a lot of the oxygen out of the room and budgets are being assigned to the large LLM providers. A lot of products are "fighting back" by calling themselves an AI product or AI first... despite just bolting on peripheral OpenAI integrations.

Per seat pricing plateau/decline

AI makes workforce/employees highly scalable, assuming you still need some level of human involvement for when the AI get stuck/go off piste, but where you used to base your SaaS product costs on human users (seats), that's going to probably plateau in a world where one marketing person can do the job of 20.

AI eats your product

The models/harnesses consume the core offering/value proposition of your app e.g. You provided market intelligence and it's emailed as a nice report at the start of each week, now people can schedule a Task or Job in Claude Cowork or OpenClaw to get this report.

Software moat is gone

An engineer with a swarm of agents will be able to replicate your product in a week/weekend. "Taste" and intimate customer understanding isn't really a moat when they can fast follow your decisions or design.

The client/agent decides to build your product

The build vs buy equation has radically shifted:

The cost/effort of building, deploying and maintaining software has collapsed.
The data that was going into the SaaS business DB will stay in their network.

You're not bespoke

This is linked to the one before, but I think it's important enough to be it's own section and people need to think about it. When you're building a SaaS product, it's a series of compromises and tradeoffs on whether you build feature X or Y, which sector and size of business to tailor it for etc... you can't do everything for everyone or you'll end up with a Frankenstein app that's too complex for most users. Now if the client builds the app themselves, they can tailor it exactly to their needs and they don't have to compromise on anything, it will work exactly how they want it to work, with the data taxonomy and integrations that are relevant for them.

Competition

Competition is going to 10x-100x in a world where someone can type a prompt into an AI system: Build me x, and it delivers a well designed and functional application.

Advertising

Linked to the competition being more ferocious, the marketing channels are also being saturated: "Personalized" (read AI) outreach flooding inboxes, agents calling phones, social media awash with AI content. There's also an open question... can you afford large human sales/marketing teams with the downward pressure on product pricing? Do humans help you cut through the agent/AI noise?

Customer base disruption

Even if you have a moat and plan to survive this storm, you also need to think about the exposure of your customer base e.g. If you sell portfolio management software for graphic designers and nano banana 5 comes out or you sell something to help human software engineers write cleaner code and Opus 5/Mythos is here.

Lower valuations

SaaS was a really reliable and predictable model which allowed investors to forecast the future revenues of the company, because of the pressures outlined above and the rapid improvement of the AI systems, that makes the future revenues a lot less predictable in this rapidly changing world.

I used the term LLM and AI interchangeably in this post.

This post was typed by hand by a human and will contain mistakes and hallucinations.