How we build now.

Ten years ago, Alexa, Polly, and Lex felt like magic. Speech to text. Text to speech. NLU. It was a big deal, and it worked.
Then came generative AI, LLMs, foundational models, third-party IVAs — and a few years of duct-taping it all together inside Amazon Connect and Lex while hoping for the best. If that era left you confused, you weren't alone.
Now we have Nova Sonic, Bedrock Agents, and native integrations that actually make sense. But here's the thing — the way we build has fundamentally changed. We're not writing contact flows. We're not defining intents and slots. We're building agents, writing prompts, selecting foundational models, and wiring up MCP servers and vector databases.
That's a lot. And it raises real questions around the three levers that actually matter in a contact center: cost, sentiment, and containment. The new approach moves all three — but how do you think about token costs, context windows, and model selection without losing your mind?
That's what this session is about. I won't pretend to have all the answers, but it took me a while to get here and I'd rather share what I've learned than keep it to myself. We'll cover the shift in thinking, get into a live demo, and hopefully leave with enough to get our hands dirty.
Virtual. One hour. June.

Oh, also, we will be using an meeting platform built in Kiro!