Big SaaS Is in Big Trouble

Here's a deal you've taken a hundred times without thinking about it. You have a business problem. Someone says, "there's a platform for that." You sign up, pay per seat per month, and let a vendor handle the rest. For about fifteen years that was the obvious move, and most of the time it was the right one.

That deal is quietly coming apart. If you run a small business, that's good news, and I want to explain why.

The deal SaaS offered, and why it worked

Be fair to SaaS for a second, because it solved a real problem. Building your own software used to be slow and expensive. You needed developers, months, and a budget most small businesses don't have. Renting a finished tool that someone else maintained beat all of that, easily.

The catch was in one word: finished. A product built for ten thousand businesses has to be generic. So you paid for a hundred features to use ten, bent your process to fit the software, and added one more login to the pile. Then you did it again, and again.

The pile is real, and it's measured. The average company now runs around 106 SaaS apps, and close to half of those licenses sit unused. That isn't a glitch. It's the natural end state of a model where the answer to every gap is to rent another generic tool.

What changed: build versus buy just flipped

The only reason renting beat building was cost and time. AI collapsed both.

I'm not speaking in the abstract. At AI-heavy teams, about a quarter of the code being shipped is now written by AI, and Google has said more than a quarter of its new code is AI-generated. On the ground that means something concrete: a person who isn't a professional developer can describe a specific need and have a working tool that fits it in an afternoon, instead of a quarter. And AI agents can handle the unglamorous part: keeping it running, watching for errors, doing the upkeep.

Once building the thing that fits exactly is cheap and fast, the case for renting something that fits seventy percent gets a lot weaker. That's the whole shift. Big SaaS spent a decade betting you'd never build your own. That bet is no longer safe.

What your stack looks like on the other side

This isn't a call to fire every vendor. Some software is genuinely better bought, and you should keep it: your accounting system, your email, the core system of record your business runs on. Buy those, commit, and move on.

It's the long tail that changes, the dozen half-used tools you bought for one feature each. That's where the future is a handful of small, purpose-built tools that fit how your team actually works, instead of generic platforms your team works around. Fewer logins. Less sprawl. Tools people use because they were built for them.

The market already senses this. After years of pile-up, companies are actively cutting tools and consolidating; the average app count peaked in 2022 and has been falling since. The treadmill is slowing down on its own.

Where this breaks

None of this is magic, and I'd be lying if I sold it that way.

When you build, you own it. There's no vendor to call at 2 a.m., the maintenance is yours, and so is the security. AI will hand you a confident answer that's flat wrong and never flinch, so a person still has to own the judgment. Your data turns into a decision you have to make on purpose: what's allowed near these tools, and what is never pasted in. And sometimes the boring truth is that the SaaS product really is better, and paying for it is the smart move.

The point was never "build everything." It's that "buy" stopped being the automatic answer.

What I'd do if I were starting today

If you're staring at a stack you're tired of paying for, a few moves that have worked for us:

• Before you renew the next contract, pull the usage numbers. If barely anyone logs in, that's your answer.
• Take one tool you bought for a single feature and ask whether you could build just that slice. You'd be surprised how often you can.
• Keep people on the judgment and let AI agents carry the upkeep. Don't flip those.
• Decide what data is allowed near any AI tool before anyone uses one, and write it down.

None of this requires becoming a software company. It requires being willing to stop renting by reflex.

Key takeaways

• SaaS won because building software was slow and expensive — which is exactly the assumption AI just broke.
• The average company runs ~106 SaaS apps with roughly half the licenses unused; that sprawl is the model working as designed, not a fluke.
• With AI, a non-developer can build a tool that fits exactly in an afternoon, and agents can keep it running, so "buy" is no longer the default answer.
• Keep the few platforms genuinely worth buying, replace the half-used long tail, and respect the real trade-offs: ownership, judgment, and data.

How we help

At Amoeba Networks we help New York small businesses make exactly these calls — what's worth buying, what's worth building, and what's worth killing — and then we run the IT and the tooling so you don't inherit a new mess. Because the moment you start building and connecting your own tools, security and data handling stop being the vendor's problem and become yours, and that's a lot of what we do.

If your stack has gotten heavier than your business, that's a good conversation to have. We'll help you find the few things worth doing, and get them done.