Human and agent experience design

The human half is explainability and trust: the interface shows why the model did what it did. The agent half is structure: states and surfaces a machine can read and operate without human oversight. Serve only the human, and agents break your flows. Serve only the agent, and people stop trusting what they can no longer follow. Product leaders keep asking the same question: our product will be used by AI agents on behalf of users, how should the UX change? Here, we'll give you the experience-backed answer.

A strategically built interface does one job exceptionally well — it shows the product at users' fingertips can be trusted. Think of sources shown next to the answer, confidence that changes the interface's behavior when the model is unsure, undo on every AI action so a mistake costs nothing, and explanation depth matched to the reader (because an analyst and an exec need different answers). The decisive moment here is the moment of doubt, when the user decides whether to act on what the model said. The interface either earns that action right there or loses it. We design for that moment, with your data, in your product.

An AI agent never sees your product. It reads it. Agent experience is the work of making your product legible to that reader. So how do you design interfaces AI agents can use? Give every element a meaning a machine can read. Give every action a name it can call. Keep both stable. And report state in terms an agent can act on (working, failed, waiting for approval). Agents are already arriving. What you control is how much they understand.

Patterns in your design system. Explainability components, agent-facing schemas, state definitions, and handoff flows land as documented, reusable pieces your team extends after we leave. The point is that the two-audience thinking survives the handover: the next feature your team designs asks what the person needs to see and what the agent needs to read.