When AI first jumped into IDEs, most of us were not ready for the role it would take in the coming years. What started as auto-complete on steroids quickly evolved into something much bigger — GitHub reports that Copilot now generates around 46% of code for its active users, Google admits that 25% of their code is AI-assisted, and Microsoft’s CEO casually mentioned that AI writes up to 30% of theirs.
We’re not talking about toy projects here, this is production code at the biggest tech companies in the world. The conversation around SaaS has gotten louder and louder, with many calling the death of the entire model. Having been in this business for quite some time, I think the reality is a bit more nuanced than that.
First: The sceptics
Many are still sceptical. Some for legitimate technical or capability reasons, some in denial, and some in pure self-preservation mode. And I get it, having skin in the game does that to everyone. But here’s the thing: it doesn’t really matter what any of us think about it. What is happening is pretty undeniable. AI in development is here, it’s a fact, and the world is moving forward regardless of anyone’s opinion on the matter (which will always happen with significant technological shifts).
What’s interesting is that it’s not just “speeding things up.” It’s doing two very separate and almost contradictory things to SaaS at the same time.
Build your own
The first and most obvious shift is that “anyone” can now build their own version of the tool they’re paying for. Customisation used to be the trade-off. You accepted the features the SaaS bundled (and the price tag that came with it) because building your own alternative was prohibitively expensive. That math has changed dramatically.
When someone only needs two features out of twenty, it no longer makes sense to pay for a subscription that bundles all of them, or a subscription altogether. The whole SaaS value proposition was built on the cost gap between “make it yourself” and “just pay for it,” and that gap is shrinking fast. A custom-built tool is local, trusted, adaptable, controllable, and can move in whatever direction you need it to. And the cost is a fraction of what it used to be, if anything at all.
Individuals and organisations with some tech chops can now shave off a part of their costs, “easily”, by creating their own solutions. Sure, they will not scale, change fast or have support, but do they really need to?
The niche explosion
The second thing is maybe even more interesting. SaaS products that were too specific, too niche, would never have been viable before due to the market size simply not justifying the investment. Now the investment to build something is one afternoon and three beers. This opens up all sorts of options that didn’t exist before, and that’s genuinely great for the world in general.
The paradox
Here’s where it gets counter-intuitive. You’d expect that if it’s so easy to build software, we’d see a massive wave of AI-generated clones replacing the big SaaS players. But we haven’t seen this to scale, and I don’t think we will. The same people who are using AI to build these things are still, in their vast majority, developers, and they know a few things that the “SaaS is dead” crowd tends to overlook.
A successful SaaS is way more than just code. One needs support, marketing, sales, channels, distribution, operations, and above all: strategy. None of the big players “launched and forgot.” They are relentless at iterating. These companies have considerable products with years of accumulated knowledge baked into them, and even if someone could clone the codebase in a day, the original would still outshine most competitors by tomorrow. AI cannot replace years of experience in strategy, planning, market knowledge, distribution channels, and partnerships.
Peter Thiel makes a related point in Zero to One: “real lasting value doesn’t come from copying what already exists, it comes from creating something genuinely new”. And that’s exactly the dynamic at play here. Building a cheaper version of someone else’s product is the same as putting the “10-minute workout” out of business with a “9-minute workout.” It’s great until the “8-minute” guy walks through the door.
All of this is deterring most people from challenging the established players. The moat was never the code.
What comes next
As it gets easier to build custom applications, these will inevitably take away market share from a significant chunk of SaaS. But I doubt these products will be replaced by other SaaS versions any time soon. They’ll be replaced by local, corporate, and personal versions of once-existing apps. The disruption won’t come from competition within the same model, it’ll come from the model itself becoming irrelevant for certain use cases. When people can build exactly what they need, they stop shopping.
Verdict
So, is SaaS in danger or not. Well, yes and no. Some of it definitely is, specially the overpriced, feature-bloated tools that survive on lock-in rather than genuine value. But the ones that have built real moats through deep integrations, network effects, data advantages, and relentless iteration will endure, at least for a while longer. The rest should be paying attention.
The true leverage right now is in adapting, knowing and listening to the audience and distribution/logistics.