/agents
Designing what AI does on your behalf.
Agents aren't a feature. They're a product decision — about what you automate, what you leave human, and how you make that boundary feel intentional rather than arbitrary.
The hardest part of agent design isn't building the agent. It's deciding what it should be allowed to do — and making that legible to the people it's working for.
Autonomy creates anxiety when users can't see what's happening or why. The experience design of an agent system is fundamentally about trust: earning it, maintaining it, and recovering it when things go wrong.
We design the human side of AI systems. The surfaces where agents report, request, and hand back. The moments of confirmation and intervention. The tone the system uses when it needs to explain itself.
We think in orchestration — not just individual interactions, but the full shape of how humans and AI systems work together over time.
What this covers:
- 01Agent UX design—The interfaces where agents act, report, and request. Transparency without noise.
- 02Human–AI orchestration—Mapping what's automated and what isn't — and designing the handoffs between them.
- 03Trust systems—How the system explains itself. Error states, confidence levels, audit trails. The design of accountability.
- 04Agent persona—The character your agent carries. Name, voice, behaviour. Distinct enough to be trustworthy. Subtle enough to stay useful.