The Interface Is Moving. Your Job Follows It.

Something changed under your team's feet this year, and it isn't the thing your feed keeps yelling about. The loud story is "learn the AI tools or fall behind." That part is real, but it's the easy part. The harder shift is where the interface lives now, what your team owns when the assistant becomes the front door, and when to stop smoothing things out. Let me catch you up.

Tool fluency is table stakes now

Start with the good news, because it's the part your team already feels. AI lets one person move fast. But speed alone isn't the win. Figma's 2026 AI report found the vast majority of designers, developers, and PMs say learning to use AI tools well is essential. Fine. That's the floor, not the ceiling.

The real gain shows up when one person's knowledge becomes something the whole team can use. A Figma researcher vibe-coded a website to host survey data so it wasn't "locked in my head or on my machine." Another team built a one-click brand-texture tool anyone could use. The instinct here is old: spot where people get stuck, remove the friction, build the tool. AI just lets more people act on it. Instead of one person moving 10x alone, the team moves 10x together.

The catch nobody puts in the roadmap

Here's what the speed hides. Figma also found that teams and leaders aren't adopting AI at the same pace: 20% say individual contributors are pulling ahead without support, and 27% say leadership is pushing AI while teams struggle to keep up. Left alone, that gap widens. The people already ahead keep gaining. Everyone else falls further back.

The economics make this worse than it looks. Exponential View lays out an AI J-curve: the learning bill arrives before the returns, so a rollout that's working can look expensive and even irrational before it looks productive. Winners and losers look identical at first. The trap they name is the "project accumulator," the company that keeps launching new AI projects but never figures out what separated the wins from the losses. Nothing carries forward. Each project starts with the same odds as the last. GM did this in the 1980s. NUMMI outperformed every other plant, GM watched it happen, and the lesson traveled too slowly to matter.

When the assistant owns the screen

Now the part that changes your job description. More tasks start in a chat box instead of your app. Patrick Neeman argues that Jakob's Law, the old rule that users expect your site to work like the ones they already know, now points at the assistant itself. People spend their time in Claude and ChatGPT, so that's the reference point they carry everywhere.

The consequence is blunt: for a real set of tasks, you don't own the surface anymore. A user asks Claude to find every auto-renewal clause in their contracts. Claude reaches your platform through a connector, does the work, and answers in the chat. The user never sees your navigation, your dashboard, or your brand. Your product did the work and stayed invisible. Your job shifts from designing the destination to being legible through someone else's.

That means retiring some scoreboard metrics. If the assistant does the task, your time-in-app and page views fall even when you're winning. Stop counting screens saved as a victory. Start mapping the end-to-end journey and marking where the handoffs to the assistant fall, because that's where expectations break.

The case for slowing users down

All this points at frictionless as the goal. Sometimes it is. Darren Yeo describes an airport ops leader who told him to "make the experience forgettable." For check-in and immigration, smooth and invisible is the gold standard. Nobody wants a memorable moment at passport control.

But Yeo's team learned the other side the hard way. They built a parents' app feed to be completely frictionless. Adoption spiked, and support tickets spiked right alongside it. Parents lost track of attachments and missed details in the infinite scroll. So they added friction back: truncated text that forces a tap to expand, prompts that make parents pause and reflect. Time in app went up, and so did satisfaction. AI raises the stakes because frictionless prompting can create passive users who take the first answer and skip judgment. Remove friction where it blocks progress. Keep it where it shapes a decision or builds trust.

The deep cut

Here's the piece that's easy to miss and actually changes your Monday. When the interface moves into a chat box, your product needs conventions spelled out, not assumed. Neeman's point is that the agent reading your product doesn't carry a lifetime of human habit and can barely remember the last conversation. So your design system and its semantic layer become the thing an agent reads to use your product well.

That reframes what your systems team ships. It's not just internal consistency anymore. It's whether your product is legible to a machine acting for a user who never opens your app. Meanwhile, don't burn energy on a rebrand of the discipline itself. NN/G surveyed 604 professionals and found "UX" still appeared in 70% of responses, with no alternative close. Renaming won't fix your pressure. Making the work legible, to stakeholders and now to agents, will.

Three questions for your team

1. Which of our top tasks now start in an assistant instead of our product, and is our platform legible enough for an agent to complete them well?
2. Are we a "project accumulator"? Name one thing a failed AI experiment taught us that carried into the next one. If we can't, that's the problem to fix.
3. Where did we strip out friction that we should put back, because a user needs to pause, understand, or decide before they act?