People keep saying “AI is the death of SaaS.” I think that’s slightly wrong.
AI alone is not killing SaaS. What’s actually happening is that AI + the thousand tiny annoyances of SaaS is killing SaaS. For years, we tolerated software because building custom software was expensive, slow, and painful. So even if a product was only 80% right for your workflow, you adapted yourself to the software instead of the other way around.
But AI changed the cost equation. At one of my companies we use Retool for reporting. Retool is great at what it does. This is not a knock on them. But our CEO used Claude to mock up the exact dashboard he wanted. Traditionally, I would have built that inside Retool. Even with Retool AI, it still would’ve taken time, compromises, and it definitely wouldn’t have matched the design exactly.
Instead, I dropped the design HTML from Claude into my normal Copilot workflow and had the exact dashboard deployed to Vercel in about 2 hours. That’s the shift. The “build vs buy” decision used to happen at massive scales. Now AI has pushed that threshold down so far that even minor annoyances start tipping the scale toward “just build it.”
I saw this personally with Log My Lift. There are already tons of workout tracking apps. Most are honestly pretty good. But every single one had little things that annoyed me:
– How exercises were entered
– How sets/reps were logged
– Handling missed reps or partial sets
– Supporting timed exercises vs rep-based vs weight+rep equally well
– How progressive overload was visualized
– Random UX friction everywhere
None of those issues individually justified building a new app a few years ago. Today? They do. Because now the question isn’t:
“Should I spend 6 months building custom software just for myself?”
It’s:
“Should I spend a weekend fixing all the things that annoy me?”
That’s a completely different calculation.
This is the new reality. We all have the tools to build what we want instantly.