We're building Ideate, an AI-enhanced creative workspace for design teams. Six months in, we added an AI agent called Sage -it analyzes uploaded content, extracts client and contact profiles from conversations, and scores feedback sentiment. On paper, it removes cognitive load. In practice, it created new cognitive load we didn't anticipate.

This talk is the honest story of fitting an autonomous AI agent into a 4-person team's workflow. I'll cover what happened when
Sage started surfacing information nobody asked for, why we had to build trust boundaries before accuracy improvements, and the interaction model we landed on — where the AI is a teammate with a defined "Team API," not a magic oracle.

If you're thinking about adding agentic AI to your team's flow, I'll share the real tradeoffs — not the vendor pitch.