Designing for no one: The UX designer's identity crisis (and rebirth)

There's a particular kind of existential dread that comes with realising your entire profession might be built on a premise that's about to vanish.

For UX designers, that premise is this: humans interact with digital interfaces. You design those interfaces. You make them usable, delightful, accessible, and effective. Your job is to understand human needs, cognitive patterns, and behaviours, then craft experiences that serve them.

But what happens when humans stop interacting with interfaces?

When AI agents do the interacting instead, clicking through your carefully designed flows in milliseconds, extracting data without seeing your layout, completing transactions without appreciating your micro-interactions, what exactly are you designing for?

Autonomous agents are here. And they're really good at navigating interfaces designed for humans. Which means they don't need you to design interfaces for them at all.

If agents bypass interfaces, what does UX design even mean?

The slow realisation that nobody's looking

I was at an event a few weeks ago and sat next to a CEO of an e-commerce company. They said they're seeing an increasing percentage of transactions that complete in under 10 seconds. A lot faster than any human could navigate their checkout flow. At first they thought it was bots. Then they realised it was early-adopter customers using AI assistants to handle their shopping.

These assistants weren't admiring the clean design. They weren't appreciating the carefully crafted product photography. They weren't reading the thoughtful copy. They were extracting structured data, comparing options, and executing purchases.

All the design work, testing, optimising layout hierarchy - completely invisible to the systems starting to use the site.

This is the future arriving in slow motion. Every interface you design is increasingly likely to be consumed by agents, not humans. And agents don't care about your design.

What agents actually need

If agents don't need beautiful interfaces, what do they need?

Structured, semantic data. Agents want to know what's on a page, what actions are available, what the relationships between elements are. They don't care if it's visually pleasing, they care if it's machine-readable.

This means proper HTML semantics, ARIA labels, schema markup, clear hierarchies, consistent naming conventions, and accessible patterns. Not because you're trying to help people with screen readers (though that matters too), but because agents parse your interface the same way assistive technologies do.

Predictable patterns. Agents learn interface conventions and expect consistency. Random creative flourishes that deviate from patterns make you harder to parse. Agents reward boring, predictable, standards-compliant design.

API access. Ideally, agents don't use your interface at all. They go straight to your APIs. If your service doesn't have clean APIs exposing the same capabilities of your UI, an agent will scrape the interface instead. And they'll hate you for it (in the algorithmic sense of deprioritising you for being annoying to interact with).

Fast, reliable execution. Agents don't care if your site is beautiful. They care if it's fast, if requests succeed consistently, if errors are clearly communicated, if rate limits are documented. Your performance budget matters more than your aesthetic vision.

Trust signals. Agents need to verify they're interacting with the legitimate service, not a phishing site or scam. This means proper security certificates, consistent brand identifiers, verifiable domains, and clear authentic flows.

What's not on the list - Visual design, delightful interactions, emotional resonance, brand personality, narrative flow, gamification, or any of the things us traditional web folk and UX designers spend most of our time on.

What survives: designing for high-stakes human oversight

Humans won't completely disappear from digital interactions. They'll remain involved in specific contexts:

High-stakes decisions. When the decision has major consequences like buying a house, choosing medical treatment, making an investment, humans will want to review what their agent recommends before executing. Someone needs to be designing these review interfaces.

Edge cases and exceptions. When something goes wrong, when an agent can't complete a task, when a human needs to intervene, you need well-designed exception handling interfaces.

Regulatory requirements. Many industries require human confirmation for certain transactions. Financial services, healthcare, legal services will all have mandated human oversight moments. These need to be designed.

Trust verification. When an agent recommends something important, humans want to verity it makes sense. You need interfaces that let humans quickly audit agent decisions and understand reasoning.

Identity and status purchases. For goods that express identity or signal status, humans will keep making their own decisions. Luxury brands, fashion, art, experimental services - these still need human-facing design.

So, UX doesn't completely die. But it contracts dramatically. Instead of designing the primary interaction path for millions of users, you're designing exception cases, oversight moments, and verification interfaces for a much smaller set of high-stakes contexts.

That's a much smaller design surface area. Which means fewer designers needed.

The split: UX becomes two distinct disciplines

I think what actually happens is UX splits into two separate career paths:

Human oversight designer. You specialise in high-stakes decision interfaces, exception handing, verification moments, and oversight flows. You're designing for humans, but in a much narrower context than today's UX. You work closely with legal, compliance, and trust & safety teams. You're as much about liability management as usability.

Required skills:

• Traditional UX research and design

• Understanding of cognitive bias and decision-making

• Regulatory compliance awareness

• Crisis communication and error handling

• Accessibility for humans in stressed states

Agent Experience Architect. You specialise in capability structure, semantic design, API ergonomics, and agent-facing interfaces. You're designing for machines, which means you work more like a technical architect than a traditional designer.

Required skills:

• Information architecture at scale

• API design and documentation

• Semantic web standards

• Understanding of how agents parse and evaluate services

• Data modelling and taxonomy design

  
  

These require overlapping but distinct skill sets. Most current UX designers are optimised for the first path but will need significant retraining for the second.

What to do if you're a designer right now

If you're reading this as a practising UX designer, you're probably wondering what you should actually do.

Here's my honest advice:

Short term (next 12-18 months):

Start learning adjacent skills. Focus on:

• Information architecture and semantic design

• API design principles

• How agents parse and understand interfaces

• Accessibility standards (because agent parsing is similar to screen reader parsing)

• Data structure and taxonomy design

Build hybrid skills. Work on projects that serve both humans and agents. Design interfaces that are beautiful for humans but also semantically clean for machines.

Medium term (1-3 years):

Choose your path. Are you more interested in:

• High-stakes human oversight (traditional UX in specialised contexts)

• Agent experience architecture (semantic design for machines)

Start positioning yourself for that path. Take on projects, build portfolio work, develop the specific skills needed.

If you choose the agent path, consider learning:

• Basic programming (to understand how agents consume APIs)

• Semantic web technologies

• How to design and document APIs

• Agent platform ecosystems

If you choose the human oversight path, develop expertise in:

• High-stakes decision design

• Regulatory and compliance requirements

• Crisis communication and error handling

• Cognitive bias and decision support

The philosophical bit: what does it mean to design when nobody's looking?

There's something profound and slightly sad about this shift.

UX design, at its best, is about empathy. Understanding human needs, frustrations, and delights. Creating experiences that feel intuitive and satisfying. Building interfaces that anticipate what people want and guide them gently toward their goals.

It's a deeply human discipline. And that's beautiful.

But in an agent-mediated world, empathy for end users becomes indirect. You're not designing for the human, you're designing for the agent that serves the human. The relationship is mediated. The feedback loop is algorithmic. The satisfaction is measured in selection rates and trust scores, not in user delight.

Some designers will find this alienating. The craft loses some of its emotional resonance when you're optimising for machine parsing rather than human joy.

Others will find it clarifying. Stripped of the subjective aesthetics and emotional manipulation, you're left with pure function. Does it work? Is it clear? Can agents understand and use it effectively? There's a certain purity to that.

I don't know which camp you'll fall into. But I know the shift is happening whether we like it or not.