The idea that the best interface is no interface has been kicking around design circles for over a decade now. AI has given it a seductive second wind — and I think we should be worried.
Because "invisible design" sounds like sophistication. In practice, it's often just opacity with better branding.
What happened
There's a renewed push in the design world to argue that traditional UI elements — buttons, menus, forms — are relics. The thesis: AI-powered systems should anticipate what users want and act on their behalf, making the interface itself disappear. Intent-driven experiences. Ambient computing. The screen fades away, and the system just… handles it.
It's a compelling vision. It's also, I'd argue, the wrong one for this particular moment.
The bigger picture
Here's the thing about invisible systems: they only work when they're right.
A thermostat that learns your preferences and adjusts automatically is lovely — until it decides 18°C is your ideal temperature in January and you're sat there in a jumper wondering why you can't find a dial. The invisible interface failed, and now the user has no way to understand why it failed, or how to fix it.
Scale that problem up to AI systems making consequential decisions — curating your content, adjusting your website, filtering your messages, generating your reports — and "no interface" starts to look less like elegance and more like a blindfold.
We've spent years talking about AI transparency. We've wrestled with labelling requirements and disclosure patterns. And now the design community is simultaneously arguing that the best approach is to hide the entire system from view? That's a contradiction we need to sit with.
The real problem with "no interface" thinking isn't that it's wrong in theory. It's that it assumes a level of AI reliability we simply don't have. Current AI systems hallucinate, make confident errors, and reflect biases in their training data. When those systems are visible — when the user can see what the AI did and why — there's an opportunity to catch mistakes. When they're invisible, errors compound silently.
And there's a deeper issue for us as designers. If the interface disappears, so does our most important contribution: making complex systems legible to the people using them. That's not a side effect of design. That's the job.
Why legibility beats invisibility
There's a useful concept in design called "seamful design" — the idea that deliberately revealing the seams of a system, the places where different components connect and where things might go wrong, actually helps users build accurate mental models of how something works.
It's the opposite of seamless. And right now, for AI, I think it's more honest.
Consider dashboard design. There's been some really sharp thinking lately about reframing dashboards not as passive data displays but as active decision-making tools. The best dashboards don't hide complexity — they make it navigable. They surface the right information at the right moment so a human can act on it with confidence.