Building Real Agents - From Loop to Voice
You can now build a voice agent that picks up the phone and remembers people across calls.
3-hour live session · 90 min theory · 90 min hands-on · plus one exercise
The 3-hour live session
Theory grounds you. Hands-on lands it. Both happen in the same Saturday session.
Theory
Agents stop being magic. They're a while loop with a prompt and tools - that's it. We separate workflows from real agents (most 'agent' content on LinkedIn is workflows). We also talk about the UX of AI products you're about to ship - chat vs voice vs background agent vs assist - because the wrong pattern kills more AI features than bad models do. By the end, every agent announcement reads differently.
Hands-on
Build something that thinks. Then build something you can call. Hang up. Call back. Ask if it remembers your name. The room gasps every cohort when this works - and it works. By Saturday evening you'll have a phone number that calls you by name. Project 2 complete.
Your exercise for the week
Give your voice agent something to remember about its callers that matters to your use case. Test with two calls. Bring the transcript.
The moment that lands
I call my own Twilio number live in front of the cohort. Hang up. Call back. "Do you remember my name?" It does. Gasps every cohort. A Cohort 1 participant implemented a version for her company within two weeks of the session.
What I'll keep saying
- “That is the agent. That is it.”
- “We're solving the most important AI limitation: it has no memory.”
- “Start simple. Only add complexity when the basic task is competitive.”
Want this week in your real life?
DM me on WhatsApp or book a 30-minute call. The cohort is 12 to 18 PMs and I personally vet every applicant.