Designing for Agents: When the User Isn't Human

About the Talk

For years, designing well has meant designing for people: how someone reads a screen, where their attention lands, what makes them hesitate or trust. Agents do none of that. They parse structure, follow rules, and act in milliseconds. Aswin Ranganathan used that gap as his starting point and asked what happens to hard-won design craft when the user is a machine working on someone's behalf.

Aswin led design at Delivery Hero before building an agentic product through his own company, Sunlight, so he came at the question from both sides: the design practice that assumes a human is looking, and the engineering reality of a system that never does.

His answer was less dramatic than the framing suggests. Designing for agents turns out to be more relatable than most designers expect, because the things an agent needs, clear structure, predictable behavior, and unambiguous state, are the same things that were always underneath good interface design. The screen was just hiding how much of it we were getting away with.

Key threads from the talk:

• ●What actually changes: the concrete differences between designing for a person scanning a layout and a system parsing structure.
• ●Why it is more familiar than it looks: how the demands agents make on a product map onto principles designers already hold.
• ●From design lead to agentic product: what building on the other side of the problem taught him about where interfaces break.
• ●Where the craft still matters: the parts of the work that do not disappear when nobody is looking at the screen.