AI and the cost of experimentation

We’re building something new at Pagecloud. One thing we’ve noticed in our own development process is that AI has dramatically reduced the cost of trying things. Not incrementally — dramatically. The kind of shift that changes how you think about building products entirely.

The old way

If you’ve worked on a small team shipping software, you know the pattern. Someone has an idea. It sounds promising. You spend a few weeks building it out, another few weeks testing and polishing it. By the time you put it in front of real users, you’ve invested enough time and energy that sunk cost kicks in hard. You ship some version of the thing regardless of what the feedback says, because throwing it away feels like waste.

The investment warps the decision. You stop asking “is this the right thing?” and start asking “how do we make this thing work?” Those are very different questions, and the second one leads you down paths you probably shouldn’t be on.

This isn’t a failure of discipline. It’s just human nature. When you’ve poured weeks into something, your brain will find reasons to keep going. Every team I’ve ever been on has done this at some point. The cost of experimentation was high enough that you couldn’t afford to treat experiments as experiments.

The new way

Here’s what it looks like now. Someone has an idea. By the end of the day — sometimes by the end of the hour — there’s a working version. Not a mockup, not a design, a working version. You put it in front of customers. You get real feedback. Then you make a decision based on actual data: throw it away, iterate on it, or double down.

The critical difference isn’t speed, though speed is part of it. The critical difference is that you haven’t invested enough to feel anchored. Throwing away a day’s work doesn’t hurt the way throwing away a month’s work does. You can be honest about what the feedback is telling you because you don’t have a sunk cost narrative running in the background.

We’ve thrown away more ideas in the last few months than we did in the previous few years. And that’s a good thing. Each one taught us something, and none of them cost enough to distort our judgment about what to do next.

Beyond developers

This is where it gets interesting. What we’re experiencing as developers — this collapse in the cost of trying things — is exactly what we’re building for non-technical people at Pagecloud.

Here’s an insight from a decade of watching small businesses interact with their websites: most of them rarely change anything. They build it once (or have someone build it for them), and then it sits there. Not because they don’t have ideas for what could be better, but because the perceived cost of making changes is too high. They aren’t experts. They don’t know what they don’t know. So they only do what they absolutely “need” to — update an address, add a holiday notice, maybe swap out a photo once a year.

That’s an enormous amount of unrealized value. Every one of those businesses has ideas they’d try if trying were free. New landing pages, different messaging, updated designs, A/B tests they’ve heard about but never run. The bottleneck has never been imagination. It’s been cost.

The leverage

AI alone doesn’t solve this. Making it marginally easier to edit a website isn’t the unlock. The real leverage comes from combining AI with domain expertise — not just making it easy to try things, but identifying the right things to try.

That’s the difference between a tool that says “here, edit your site” and one that says “based on what’s working for similar businesses, here are three things worth testing on your homepage — want me to set them up?” The first one still requires the business owner to know what to do. The second one collapses the expertise gap and the execution gap at the same time.

This is what we’re building. AI that doesn’t just reduce friction but actively surfaces opportunities. The combination of low-cost experimentation and intelligent suggestions is what turns a static website into something that continuously improves.

The bottleneck is the cost of trying

Every industry has a version of this story. The cost of experimentation is the bottleneck — for developers, for small businesses, for anyone trying to figure out what works. AI removes it. Not by making you faster at the thing you were already doing, but by making it feasible to try things you never would have attempted.

We’re pretty excited about what this means for the people we’re building for. If you’re curious about what we’re working on, visit pagecloud.ai and sign up for early access.