LLMs taught users they can describe an outcome in plain language and watch software act on it. Expectations shifted fast.
Once I tried it, I wanted it almost everywhere. It’s not a new idea, but it works well enough now to stick. Instead of picking an Instagram filter, I say how I want the photo to look. Instead of keyword hunting on Pinterest, I describe the room I’m imagining and get options that feel right. On Amazon, I might not know the product name, but I can describe the job to be done and expect to find it. I tell the app what to do, and it does the thing.
Most software predates this shift, so the old UI is still the base. What’s new is that intent starts the flow. Menus and forms still matter, but they sit nearby for when you want to refine or double-check. If a product makes me learn its UI first, it kind of feels dated now.
Quality is improving, and intent doesn’t feel like a demo anymore. It works best when you can describe what you want better than you can click your way to it. It helps less when the flow has to be fixed and audited.
This doesn’t replace craft. It raises the bar. Data modeling, permissions, and latency still decide whether any of this actually works. And the obvious risks are still there: hallucinations, prompt injection, cost. You still need someone watching.
Anyway, I think it’s a great time to build. New products can launch intent-first. Existing ones can expand what users can do without burying them in UI. It’s already happening across tools I use every day. It’s on its way to being the baseline.