Radhika Advani, Google
Google logo
WorkshopMembers only

From Typing to Talking: Give Your AI Agent a Voice

Jan12
Tuesday, January 12, 2027
9:00 AM – 9:45 AM (America / Los Angeles)
Live on Zoom
Screen-share demo · 45 min

Facilitated by

Radhika Advani

Senior Product Manager at Google

About event

Every AI tool these members have built this year takes text in and gives text back. This one listens.

Radhika Advani, Senior Product Manager (ex-Google), starts with a working AI tool that only accepts typing. Live, using Deepgram and Claude, she adds a voice layer to it, then handles what breaks when a real human talks to it instead of a keyboard. One tool, one voice layer, screen shared the entire time.

You follow along and leave with a voice layer you can wire into something of your own.

What you’ll walk away with:

  1. Voice is a layer, not a rebuild. Adding it is mostly wiring, not re-architecting. You will see what actually has to be in place for a tool to take speech, and how much of it is plumbing you can reuse.
  2. Speech input breaks assumptions text never did. People ramble, stop halfway, and say things your prompt was never written for. You will see what that does to an agent and what to put in front of it.
  3. Latency is the product decision. With text, a slow response is a wait. With voice, it is a broken conversation. You will see where the time actually goes and what you trade to get it down.