Every few years, someone confidently announces the death of SaaS.
Open source was supposed to do it.
Low-code was supposed to do it.
No-code was definitely supposed to do it.
Now it’s AI’s turn.
The latest argument goes something like this:
“Code is now cheap. Companies can build their own internal tools. Therefore, SaaS as a business model is finished.”
It sounds smart. It feels inevitable.
It’s also missing the point.
AI isn’t dismantling SaaS.
It’s dismantling mediocre SaaS.
And that’s an important distinction.
The Mistake: Confusing Cheaper Code with Lower Value
Yes, generative AI has made writing code dramatically cheaper.
But SaaS businesses were never defensible because code was expensive. If that were true, GitHub and Stack Overflow would have wiped out the industry years ago.
What actually creates value in SaaS has very little to do with typing characters into an IDE:
- Workflow ownership
- Deep integration into how teams actually work
- Data gravity
- Switching costs
- Compliance, security, and trust
- Distribution and brand
AI touches some of these.
It doesn’t erase them.
Cheap code doesn’t equal cheap software. It just lowers the barrier to starting badly.
“We’ll Just Build It In-House” (Until Reality Shows Up)
Many companies are experimenting with building internal tools using AI. Some will succeed.
Most won’t.
Building version one is trivial. AI is great at that.
Living with version three, five, or ten is where things break.
Internal tools quietly accumulate hidden costs:
- Maintenance when the original builder leaves
- Security and compliance gaps no one owns
- Edge cases that only surface under scale or stress
- Lack of documentation and onboarding pain
- Opportunity cost of senior people becoming accidental software vendors
These costs don’t show up on a SaaS invoice, so they’re easy to ignore. But they’re real, and they compound.
That’s why, historically, companies outsource undifferentiated complexity. SaaS exists because most businesses don’t actually want to be software companies.
AI doesn’t change that instinct. It just delays the reckoning.
The Myth of “Outcomes Over Dashboards”
Another popular claim is that AI agents will replace SaaS products because people want outcomes, not dashboards.
That sounds new. It isn’t.
Every SaaS pitch for the last 15 years has promised outcomes. The dashboard was never the value. The workflow was.
Here’s the uncomfortable truth:
AI agents still need systems.
They need:
- Structured data
- Permissions
- Audit trails
- Integrations
- Reliability guarantees
- Clear ownership
In other words, they need platforms.
Agents don’t replace SaaS. They get embedded inside it.
The UI may disappear. The subscription might change shape. The value capture doesn’t.
Proprietary Data Favors SaaS, Not DIY
If proprietary data is king, SaaS incumbents are in a stronger position than most internal tools.
Well-run SaaS platforms sit on:
- Years of structured, permissioned data
- Normalised workflows across thousands of companies
- Patterns individual businesses can’t see from inside their own walls
Internal tools, by contrast, are often:
- Fragmented
- Person-dependent
- Poorly documented
- Hard to evolve safely
AI amplifies the value of good data. It doesn’t magically fix bad foundations.
What Actually Dies in the AI Era
Let’s be clear: some SaaS will die.
Specifically:
- Generic horizontal tools with shallow differentiation
- UI-first products with no workflow depth
- “Nice-to-have” dashboards that never owned a decision or action
AI is ruthless at exposing software that was already weak.
That’s not the end of SaaS. That’s the market working.
What Thrives Instead
The SaaS companies that win in the AI era will look different, but very familiar:
- More vertical and opinionated
- Deeply embedded in workflows
- Less about features, more about ownership of outcomes
- SaaS with agents, not SaaS replaced by agents
The bar goes up. The category doesn’t disappear.
SaaS Isn’t Dead. It’s Growing Up.
Every platform shift creates the same illusion:
that tools disappear, when in reality value just concentrates.
AI doesn’t end SaaS any more than cloud ended software companies. It changes who deserves to exist.
If anything, this is good news.
For builders who understand real workflows.
For operators who value durability over demos.
For job seekers who want to work on products that actually matter.
SaaS isn’t being eaten.
Bad SaaS is.
And honestly, it’s about time.
