Content design for AI agents

As we move from static chatbots to autonomous agents, the role of the content designer is shifting from “writing responses” to “designing cognition.” Agents don’t just talk; they reason, use tools, and make decisions. This session goes deep into the mechanics of agentic workflows — from defining the router logic that directs user intent to building structured response schemas that maintain brand integrity across non-deterministic outputs. We’ll move past the “chatbox” UI to look at the rigorous content systems required to keep agents helpful, harmless, and on-track.

You will learn how to move beyond “vibes-based” prompt engineering into the engineering-adjacent work of defining agent personae and influencing technical stakeholders on model parameters. We’ll explore how to add structure to fluid responses and, most importantly, how to move from subjective feedback to rigorous evaluation using content-led rubrics.

Whether you are currently shipping AI features or preparing for the transition, you’ll leave with a technical blueprint for designing agents that aren’t just conversational, but truly capable.

In this session, you’ll learn how to:

• Design for reasoning: Learn to define “router” logic that determines when an agent should act, clarify, or hand off to a human or another agent.
• Implement structured output: Master the use of system instructions and JSON-based schemas to ensure LLM responses follow a predictable structure without losing their natural flow.
• Operationalize evaluation: Build a multi-dimensional rubric to measure accuracy, tone, and safety, moving your team from “this feels off” to data-driven content iterations.
• Engineer partnership: Gain a “Developer-to-English” vocabulary to influence technical decisions on model temperature, context windows, and latency trade-offs.