← Browse

World's Fair 2026

Stop Writing Tone Instructions. Layer Them. - Isadora Martin-Dye, Isadora & Co

Isadora Martin-Dye , Isadora

Overview

This talk introduces a four-layer architecture for managing AI agent behavior, particularly focusing on brand voice and user interaction. The core thesis is that a single system prompt is insufficient for complex AI applications, especially where nuanced communication is critical. Instead, layering distinct responsibilities—immutable identity, situational mode, example-anchored voice, and post-generation veto—provides a more robust and reliable system. This approach moves beyond simple prompt engineering to a more structured systems engineering methodology.

Who should watch

Key takeaways

Notable quotes

*I'm not programming a robot, I'm managing a brilliant intern with an incredibly high IQ and a terrible EQ.*
*The first three layers are all instructions. Identity, conditions, and voice. They're all things you tell the model, and the model usually listens. Usually, they are a request. The fourth layer is not a request. It is reading what actually came out and decides whether to allow it to leave your business.*

Watch on YouTube →

Unofficial community note. Prefer the recording for nuance.