Designing AI That Works for Everyone: What Usability Still Has to Teach Us

We often treat AI like it's rewriting the rules of interface design - but in many ways, it’s just reminding us why the old rules still matter.

In Usability of Interactive Systems, we’re taught that simplicity takes work, trust is fragile, and universal usability isn’t optional. These lessons apply more than ever in the age of generative AI.

 AI Isn’t Intuitive - We Have to Design It That Way

Most AI systems aren’t inherently usable. They’re powerful, yes, but also unpredictable, opaque, and overwhelming.

Designing for clarity and control is now a core responsibility. That means:

Explaining what the AI is doing, and why.

Letting users undo, override, and adapt the system.

Avoiding “magic” behavior that changes contextually without warning.

 Trust Is the New UX Currency

A single confusing AI response can break user trust, and it’s hard to win back. Reliability isn’t just about uptime anymore. It’s about:

Predictable interactions.

Transparent logic.

Respect for user intent and consent.

Especially in sensitive spaces like healthcare, education, or finance, trustworthiness is usability.

 Universal Usability Is a Baseline, Not a Bonus

Too often, AI is designed for “average” users - forgetting that average doesn’t exist. People differ in:

Device access (low-end phones, slow connections).

Abilities (visual, cognitive, physical).

Languages, cultures, and digital literacy.

Inclusive AI systems consider all of this. They support screen readers. They offer visual and text alternatives. They explain outputs in plain language. They adapt to users, rather than expecting users to adapt to them.

 Power vs. Simplicity? Design for Both

We often think we have to choose: make it easy for beginners or powerful for experts. But well-designed systems support both - through:

Layered complexity (basic view vs. advanced tools).

Tooltips, previews, and progressive disclosure.

Customization without chaos.

 Don't Guess - Watch, Test, Iterate

As AI evolves, so must our testing. It’s not enough to track clicks or completion time. We need to:

Observe how users interpret AI outputs.

Measure how trust changes over time.

Design with diverse edge cases in mind - not just power users or engineers.

Final Thought

If AI is going to shape how we learn, work, and connect - it has to work for all of us. That means designing with the same care we’ve always needed: simplicity, transparency, inclusivity, and empathy.

Old usability wisdom isn’t outdated. It’s our best guide forward.