Your Next User Isn't Human: Designing for an AI-Powered World

Remember Web 1.0? Back when Information Architecture (IA) was king. Before UI kits, design systems, and Figma, we had structure. Hierarchies, taxonomies, metadata – the unseen scaffolding that made the internet usable. That's where I started. I spent my days grappling with developers over the use of MySQL with Perl or Coldfusion, arguing about naming conventions, and wading through endless sitemaps and content logic spreadsheets. Not exactly glamorous, but it was the backbone of my early UX career, before they even called the role that.

The early days were all about Information Architecture (IA). Then came the flashy era with a focus on micro-interactions, mobile-first design, and design thinking workshops. IA was quietly absorbed by UX, becoming a part of "content design" or "content strategy".

But guess what? It's time for IA’s comeback.

Because your next user isn't human.

The Rise of the AI Assistant

Recently, I watched ChatGPT’s Operator try to perform a basic task in a web interface. It stumbled. It clicked the wrong thing. It couldn’t find the button. It felt like watching a hyper-intelligent toddler trying to book a flight without a map. Not because the design was broken—but because it was never designed for an AI to begin with.

And yet, here we are: AI assistants are starting to perform real tasks across real interfaces. Paying your utility bills. Picking plane seats. Buying sneakers. Booking trips. All of this is already happening—with tools like AutoGPT, Operator, and others quietly moving through the web on our behalf.

But our UIs are not ready. They’re designed for visual processing, not machine parsing. Hidden menus, unlabeled inputs, non-semantic HTML, visually clever but structurally chaotic layouts.

That’s not a design flaw. It’s an architectural one.

IA: The Blueprint for Machine-Ready Interfaces

Information Architecture isn’t just a throwback—it’s a mirrored reflection of AI. Where AI tries to interpret and act, IA lays down the structure that makes that possible. One thinks. The other maps.

Information Architecture isn’t retro. It’s revolutionary again.

It’s about creating structure that both humans and machines can navigate:

• Taxonomy: Classifying content into clear categories that a bot can interpret (e.g., "Shipping Info," "Payment Options")

• Labeling Systems: Using consistent, semantic naming that can be inferred programmatically

• Navigation Design: Structuring journeys that aren’t just intuitive for people, but logical for agents

• Tagging & Metadata: Enabling retrieval, filtering, and recognition across contexts

In short: IA gives meaning to what design makes beautiful.

Accessibility: The Accidental Blueprint for AI Readiness

Here’s where it gets interesting.

Teams who’ve invested in accessibility are already halfway to AI compatibility. Screen-reader-friendly experiences use semantic HTML. They rely on ARIA labels. They prioritize clarity and consistency over cleverness. All of that helps AI, too.

Accessibility wasn’t intended for bots. But it turns out that making your interface usable for a blind human also makes it legible to a machine. That’s a powerful overlap—and it gives teams a head start if they’re paying attention.

Interestingly, this same principle applies in more extreme usability tests—like the “boozeability” experiments, where researchers tested website usability on intoxicated users. The logic? If a user can successfully navigate your site while tipsy, your design is probably foolproof. And it worked: the simpler the structure, the better the outcomes. Clear IA and intuitive flows survived the cognitive chaos. In some ways, designing for AI assistants is a sober version of the same idea—you’re optimizing for a user that doesn’t handle nuance, inference, or ambiguity well.

What UX Teams Should Do Now

If you’re a designer, strategist, or PM, and you’re not thinking about this shift yet—start now. Here’s where to begin:

1. Treat IA as a First-Class Discipline Again Hire or retrain for it. Audit your product hierarchies, taxonomies, and labels. It’s not overhead—it’s strategy.

2. Design with Machine Parsability in Mind Ask: Could an AI assistant understand what this button does? Could it complete this flow without guessing?

3. Double Down on Accessibility If it works for screen readers, it probably works better for bots. Make this your baseline, not your checkbox.

4. Prototype AI-Agent Use Cases Try running a few core user journeys through an AI assistant. Where does it fail? What gets lost in translation?

The Future Isn’t Human-Centered. It’s Human+Agent Centered.

This isn’t about replacing people. It’s about acknowledging that human users will increasingly delegate tasks to AI. If you care about UX, you have to care about how those agents interact with your interface.

We’re entering an A2B2C2A world: Agent to Business to Consumer to Agent.

Websites are evolving into dynamic tools that contribute to AI decision-making. For AI to access website content effectively, websites need clear information architecture, intuitive navigation, consistent design, trust, data accuracy, availability, and security.

So yeah, it’s time to bring back IA.

Not because it’s nostalgic, but because it’s essential. If your interface can’t be parsed by an AI assistant, it’s not future-ready—it’s a risk. IA is no longer just a usability layer. It’s infrastructure. It’s strategy. It’s how your product stays relevant in a world where humans increasingly delegate decisions to machines.

This isn’t about going backward—it’s about building forward, with clarity and intent. IA is the key to making sure both people and the agents they rely on can move through your experience without friction.

That’s not just good UX. That’s survival.