Designing an Agent: The Blue Print

About the Talk

Aviad Cohen, an AI Experience Designer with more than five years in B2B SaaS, walked through a practical blueprint for designing AI agents: a 30-minute talk followed by 15 minutes of Q&A, run as an online session over Google Meet.

His framing: as agents take over interactions that used to be UI-driven, designers stop designing screens and start designing behavior. The job shifts from UX to AX, Agent Experience.

Key topics covered:

• ●From UX to AX: what changes when the interface is a conversation and the product is an autonomous behavior, not a screen flow.
• ●System prompts as design artifacts: treating the prompt as a first-class design deliverable, not engineering plumbing. It defines tone, scope, refusal behavior, and the agent's worldview.
• ●Training agents through conversation: using feedback loops in real dialogue to shape the agent over time, instead of one-shot specification.
• ●The four trust stages: how user trust in an agent develops in stages, and which design moves earn (or break) trust at each stage.
• ●Pre-design questions for agents: the questions to answer before you design anything, including who the agent is for, what it is allowed to decide on its own, and where the human stays in the loop.