The Invisible User: Designing for AI Agents

For years, UX has been built around one assumption: a person is navigating the experience. They click, search, compare and decide. The system responds. That model is starting to change.

Across financial services and enterprise operations, we are seeing AI agents take on more responsibility, not just answering questions, but interpreting intent, gathering information, applying rules and completing tasks on behalf of users.

This is Agent Experience (AX) and it requires a different design mindset.

AX is not about designing better AI interfaces. It’s about designing systems that AI agents can operate reliably and safely alongside people. That distinction matters.

Most organizations still treat AI as a layer added onto existing experiences. What we are seeing instead is that AI changes the structure of the experience itself. The interaction becomes less important than the outcome.

What We Are Seeing in Practice

One thing that has stood out in client engagements is that the biggest challenge usually isn’t the AI model. It’s the workflow underneath it. Processes that humans navigate through instinct or experience often break down when an agent tries to execute them programmatically.

We have seen situations where:

• business intent was not clearly defined
• exception handling relied on manual judgment
• different systems interpreted the same action differently

From a UX perspective, these are manageable friction points. From an AX perspective, they become blockers.

The organizations moving fastest are not necessarily the ones with the most advanced AI. They are the ones simplifying workflows and structuring decisions so agents can execute them consistently.

A Pattern We Are Seeing

In one insurance engagement, the original goal was to improve digital claims intake.

The first instinct was a traditional UX approach: simplify forms and reduce steps. But during discovery, it became clear the bigger opportunity was reducing how much users needed to do in the first place.

The redesigned flow introduced AI-assisted claims handling that could pre-populate information, interpret uploaded documents, validate missing data and route claims automatically.

What changed the outcome was not a better interface. It was designing the workflow so an agent could operate within it effectively.

That reduced both customer effort and operational handling time.

Three Things We Recommend

1. Design Around Intent, Not Screens Focus on the outcome the user is trying to achieve, not the interface they move through.

2. Make Decision Logic Explicit If business rules only exist inside manual processes, agents won’t scale effectively.

3. Keep Humans in the Loop The strongest AX models are not fully autonomous. Users still need visibility, control and the ability to intervene when needed.

Looking Ahead

We are still early in this shift, but the direction is becoming clear.

The next generation of enterprise experiences won’t be defined only by better interfaces. They will be defined by how effectively systems can understand intent, make decisions and execute actions across increasingly complex environments.

In many ways, this is a shift toward designing for actions, not clicks.

That’s the shift from UX to AX.