One Agent, Many Voices

One conversational agent runs the front door of my consulting business. There isn’t a single form on the site. Depending on which door someone comes through, Cyris reads my calendar and books an appointment, conducts an operations diagnostic, works through an AI-readiness audit, or handles intake for a paid session.

In this presentation at the AI Practitioners Summit, I walk through how I built those voices and what each one taught me. An operations consultant has to hold its judgments through a real diagnostic interview. An AI-readiness analyst has to reconstruct the customer’s actual process with them, because most people can’t tell you their process timings in the abstract. The work each voice is responsible for shapes the conversation.

Underneath those professional hats is one agent loop and one hardened tool layer. The personas are versioned files. Adding a voice doesn’t require another codebase or another agent; it requires understanding the job well enough to give that voice useful instructions and the right tools.

The Cyris project brief goes further into the architecture and delivery decisions behind the system. You can also meet Cyris at Pyris Consulting.