Genesys Agentic Virtual Agent Powered by LAMs for Enterprise CX

Mar 12, 2026

Mike Szilagyi, SVP and GM of Product Management at Genesys, leads AI-driven customer experience product strategy. He discusses agentic virtual agents built on Large Action Models that plan and execute multi-step tasks. Topics include LAMs versus LLMs, real-world rebooking workflows, enterprise governance and auditability, and partnerships that make autonomous actions reliable.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

INSIGHT

LAMs Turn Conversation Into Reliable Execution

Large language models excel at generating words and understanding intents but struggle with executing multi-step tasks reliably.
Genesys built Large Action Models to plan, map API semantics, verify outputs, and connect intent to successful actions across systems.

INSIGHT

Outcome Focused Agents Resolve Problems

LAMs focus on outcomes not just answers, enabling end-to-end rebooking or fulfillment rather than giving instructions.
The model understands APIs, semantic inputs, success states and documents interactions to meet business policy.

ADVICE

Start With Governance To Trust Autonomous Actions

Design governance first: build model cards, audit logs, explainability and guardrails into the platform rather than bolting them on.
Genesys follows ISO 42001, creates model cards for each model, and traces inputs, outputs and decision paths.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

Have you ever contacted customer support with a simple request, only to find yourself trapped in a loop of scripted chatbot responses that never actually solve the problem? It's an experience many of us know all too well.

AI has made customer service more conversational over the last few years, yet there is still a gap between answering a question and actually resolving an issue. That gap is exactly where today's conversation begins.

In this episode of Tech Talks Daily, I spoke with Mike Szilagyi, SVP and General Manager of Product Management at Genesys Cloud, about a new chapter in AI-powered customer experience. Genesys has announced what it describes as the industry's first agentic virtual agent built on Large Action Models, or LAMs. While Large Language Models have dominated the conversation around AI for the past few years, they have largely focused on generating responses, retrieving knowledge, or answering questions. What they have struggled with is execution.

Mike explained how Large Action Models take the next step. Rather than simply telling a customer how to solve a problem, these systems can plan and execute the steps needed to complete a task. Imagine contacting an airline after a sudden flight cancellation.

Instead of navigating multiple menus or repeating information to a human agent, an agentic virtual assistant could understand your situation, check alternative flights, apply airline policies, and complete the rebooking process across several systems. In other words, the AI moves from conversation to action.

We also explored how Genesys approached the design of this technology with enterprise governance in mind. From explainable decision paths and audit logs to guardrails that ensure every automated action can be traced and understood, the goal is to make autonomous AI trustworthy inside complex organizations. Mike also shared insights into Genesys' partnership with Scaled Cognition and how integrating specialized models helps deliver reliable execution in real-world customer service environments.

Perhaps most interesting was our discussion about the human role in this evolving contact center landscape. As automation begins to handle routine and multi-step workflows, human agents are free to focus on situations that require empathy, judgment, and expertise. That shift raises interesting questions about how organizations design customer experiences in the years ahead.

So how will customers respond when virtual agents move beyond answering questions and begin resolving problems on their behalf? And once one brand delivers that experience, will it quickly become the expectation?

Useful Links