I've been going back to something I used to care about a lot when I was deeper into design. Not features. Not scale. But the basic things that somehow don't come up enough anymore.
How many steps does it take for the user to get something done? Where do they pause? Where do they hesitate? At what point do users feel like they've lost control of what's happening? And more importantly, what does the experience feel like when users use it for the first time, without anyone explaining it to them?
These used to be the questions. Now the conversation is different. It's about how quickly something can be built. How much of the workflow can be automated. How fast can we move from problem to something that looks like a solution.
And to be fair, AI is very good at that. You can get to something functional in a fraction of the time it used to take. You can generate flows, code, edge cases, and even documentation without breaking much of a sweat.
But when I started looking closely at products built heavily this way, something felt off. Nothing was obviously broken. Everything technically worked. But the experience felt like it was put together in a hurry. Not rushed in effort, but rushed in thought.
The system knew what to do. It just didn't feel like it understood who it was doing it for.
A familiar pattern
When manufacturing scaled during industrialisation, products became cheaper, faster to produce, and widely accessible. But they also became more standardised. Less opinionated. Built to function reliably, not necessarily to create an experience. That wasn't a flaw. It was a trade-off.
A few mature brands didn't reject machines. They just understood something most people miss — not every part of a product should be optimised for speed. They automated what didn't define the experience. They kept human craftsmanship where it actually mattered. That's what allowed them to scale without becoming forgettable.
We're at a similar point in software
AI is industrialising how we build. It's removing friction that used to slow teams down. But it's quietly pushing a lot of products toward the same shape: functionally correct, reasonably complete, but not particularly thoughtful.
And there's an important difference from traditional industrialisation. Machines in factories were predictable. AI doesn't behave like that by default. It can generate strong outputs, but it doesn't naturally carry context, intent, or consistency unless someone is actively shaping it.
Which means the human role hasn't reduced to just reviewing and approving. You're still deciding what matters, what doesn't, what needs to be simplified, and what needs to be held back.
What separates the two categories
From a business point of view, we're going to see a familiar pattern. A lot of companies will lean into speed — they'll adopt faster, cheaper software because it's available and good enough. For many use cases, that will be completely fine.
But over time, the trade-offs won't show up in whether the system works. They'll show up in how it feels to use. How quickly people understand it. How much they trust it. How long they're willing to stick with it.
Most software will become cheaper and more replaceable. A smaller category will become more expensive and more trusted. The gap between the two will widen.