AI-powered tools now let teams generate layouts, wireframes, and even full page designs faster than any human review process can keep up with. That speed creates a real problem: unchecked design debt, shallow user research, and inherited workflows that no longer fit. Across five industry reports this week, a clear pattern emerges — the role of UX in web design is shifting from creator to curator, and agencies that fail to adapt will ship inferior work. Here’s what our team is acting on right now.

Key Takeaways

• AI generates designs faster than UX teams can evaluate them, creating a growing backlog of design debt.
• AI-assisted empathy maps are no substitute for real user research with real people.
• New browser APIs like Document Picture-in-Picture open up fresh possibilities for interactive web widgets.
• Software’s role in society is shifting from tool to intermediary, changing how we should approach design decisions.
• Legacy UX processes inherited from mid-century advertising are breaking down under AI-era demands.

Design Debt Is Piling Up — Build Shared Judgement First

AI lets any team member spin up a plausible-looking interface in minutes. The catch: plausible isn’t the same as usable. Nielsen Norman Group’s latest analysis describes this as the custodial era of UX, where teams must clean up after AI. The article argues that UX professionals need to shift towards building shared evaluation criteria across the entire team, not just reviewing outputs after the fact.

We’ve started implementing exactly this for client projects. Before any AI-assisted design sprint, our team now establishes a documented set of usability benchmarks. Every generated layout gets scored against those benchmarks before it reaches a client review. It adds a small amount of time upfront. It saves significant rework downstream.

AI Cannot Fabricate Genuine User Empathy

There’s a temptation to feed a prompt into ChatGPT and ask it to produce an empathy map. The output looks convincing. It’s also largely fictional. As NN Group makes clear in their report on why AI can’t replace real research in empathy mapping, AI can organise research you’ve already collected, but it cannot generate evidence about the specific, messy experiences of actual users.

Our position: we use AI to synthesise and sort interview transcripts and survey data. We never use it to invent user personas from scratch. The distinction matters. Clients who skip real research end up designing for imaginary people.

Document Picture-in-Picture API Opens New Widget Possibilities

On the technical front, the Document Picture-in-Picture API is gaining traction as a way to build persistent, interactive web widgets that float above the main page. CSS-Tricks published a detailed walkthrough on creating web widgets using the Document Picture-in-Picture API, showing how developers can inject full HTML, CSS, and JavaScript into a standalone window.

Practical applications we’re exploring for clients include:

• Persistent live chat widgets that survive page navigation
• Floating video players for product demos
• Always-visible booking or enquiry forms on service pages

Browser support is still limited, but Chromium-based browsers cover a large share of UK web traffic. We’re building progressive enhancement layers so these features degrade gracefully.

Software as Intermediary, Not Just Tool

A thoughtful piece on UX Collective examines the function of software in society — framing it as a story of inputs, outputs, and the systems between. For our design work, this reinforces a principle we apply daily: every interface sits between a person and an outcome. The job is to shorten that distance, not decorate it.

Legacy UX Workflows Need a Hard Reset

The rituals of the 1960s advertising agency — the pitch room, the campaign cycle, the account manager handoff — still shape how many UX teams operate. A sharp analysis on UX Collective argues that Sterling Cooper built your UX process, and it’s time for a reset. Linear, presentation-heavy workflows collapse when AI can produce dozens of viable options in a single afternoon.

We’ve moved towards continuous, iterative design reviews rather than big reveal presentations. It keeps decision-making tight, reduces wasted cycles, and lets us integrate AI outputs without losing quality control.

The web design industry in 2025 is being reshaped by speed — the speed of AI generation, the speed of new browser capabilities, and the speed at which outdated processes become liabilities. Agencies that treat AI as a shortcut to skip research and evaluation will accumulate debt they can’t repay. Our approach remains straightforward: use AI where it accelerates good decisions, and keep humans firmly in charge of judgement.

Frequently Asked Questions

What is UX design debt from AI-generated layouts?

UX design debt occurs when AI produces interfaces faster than teams can properly evaluate them for usability. Over time, these unchecked designs accumulate problems that become expensive to fix after launch.

How do web designers use AI for empathy mapping without compromising research quality?

AI works well for organising and categorising user research data you’ve already gathered through interviews or surveys. It should never be used to fabricate user insights, as the output will lack the specificity of real human experience.

What is the Document Picture-in-Picture API used for in web design?

It allows developers to create floating widget windows containing full HTML, CSS, and JavaScript that persist outside the main browser tab. Practical uses include always-visible chat tools, video players, and booking forms on commercial websites.

Why do traditional UX workflows fail in AI-driven web projects?

Legacy processes rely on linear, presentation-heavy review cycles that can’t handle the volume of design options AI produces. Replacing them with continuous, iterative evaluation keeps quality high without slowing delivery.