On The SaaS Apocalypse
It's hard to make predictions, especially about the future. - Yogi Berra
Right now, my LinkedIn feed is full of hot takes about what's coming for businesses that sell software as a service (SaaS). Maybe they're doomed, because everyone can just vibe-code their own versions of task managers, calendar apps, schedulers, CRMs etc. Maybe they'll be fine, because after the initial excitement wears off, people realize that it's hard work to build and maintain a useful tool, and anyway, all this AI coding stuff is going to go away soon, hmph!
Everyone has their well-reasoned take, but I find it impossible to accurately extrapolate what will happen to software companies, because we're looking at multiple develops with different effects on software, and where things end up depends on the relative strength, over time, of these effects.
Lowered Barrier
It is true that the barrier to creating (semi-)useful things has been lowered. I once needed to extract a lot of data from a particular website and had Claude build me a custom tool for that job in about thirty minutes, give or take. But the second-closest alternative to that agent-coded tool wasn't to buy a SaaS subscription. It was not to bother.
On that front, I'm confident we'll see an explosion of custom, purpose-built tools where nothing would have been done before.
They Have AI Too!
If I really wanted to, I could spend some time and make my own to-do app. It's mostly database stuff, really, with an opinionated interface that lets you add, edit and complete tasks. With AI, I could even do it quite quickly. But guess what? The company that build my to-do app of choice has access to AI, too, so their delivery of features and their quality of maintenance will be higher, too. That app costs me about $50 a year so in order to be cost effective, I'd have to limit my time spent on my custom to-do app to not even a full hour per year.
Who Wants To Deal With Maintenance?
Lots of posts talk specifically about replacing a scheduling-tool subscription, such as Calendly, with a vibe-coded tool. Which is funny, considering that there's already a completely free and heavily customizable open-source alternative out there. https://github.com/calcom/cal.com
You just have to follow their steps for installation, set up the requirement infrastructure, host it, and you're good to go. Or you pay them $15/month on the basic plan and you're good to go. Lots of SaaS companies follow this model: Open source so you can use it for free if you know how to and feel like doing it, or have them host it and deal with all that.
The fact that this is a viable model suggests that not having to deal with everything that comes after you already have the source code is worth real money.
What About Seemingly Overpriced Enterprise AI?
First question: Is it really overpriced, or are the LinkedIn experts missing something, resulting in some "How hard can it be?" hubris? Remember, people aren't paying for the existence of code. They're paying for everything that comes after, as well. In enterprise products, that includes all sorts of security and compliance audits. You can't just tell Claude to "go make it SOC-2 compliant." (Not yet anyway. Fingers crossed.)
But fair enough, I can see a world where large companies with underutilized engineering teams let them loose to eliminate their current SaaS spend. Whether or not that makes economic sense remains to be seen. There are opportunity costs, after all. Shouldn't those engineers make the core offering more compelling?
Conclusion
We have to accept that we can't predict what exactly will happen to the subscription-model of software, but I want to leave with two concrete takeaways:
AI-coded tools are a great alternative to doing nothing
Beware the siren call of vibe-coding away your SaaS spend; put that same energy into your actual offering instead.
