Agentic CX: A Practical Roadmap for Multi‑Agent Orchestration

Write your awesome label here.

Legacy NLU assistants and basic RAG bots aren’t built for today’s AI landscape. In this webinar, Olivia Marrese (66degrees) shares a practical roadmap for migrating to modular, multi-agent AI systems, covering LLM readiness, RAG assessments, specialized agents, and how to measure success through real self-serve completion.

Most conversational AI systems today were built in a very different era.

They rely on brittle NLU pipelines, rigid intent trees, and single assistants trying to do everything at once. When generative AI arrived, many teams simply layered RAG on top and hoped it would solve the problem. In reality, this often created systems that are harder to scale, harder to govern, and harder to maintain.

So how do you migrate to a more flexible, scalable architecture?

In this webinar, Olivia Marrese, Conversational Architect at 66degrees, shares a practical roadmap for evolving legacy conversational systems into modular, multi-agent AI architectures.

We’ll start by looking at LLM readiness: how to audit your current assistant to determine what should remain deterministic, what can move to generative flows, and where specialized agents can add the most value. Olivia will walk through a structured RAG readiness assessment, helping teams identify gaps in data sources across websites, CMS platforms, and internal documentation before attempting large-scale AI upgrades.

From there, the session explores how to move toward a hybrid multi-agent model. Instead of a single overloaded assistant, you’ll learn how to design a coordinated system of agents with clear responsibilities, from routing and orchestration to domain expertise and transactional execution. This modular approach allows teams to add new capabilities without constantly rebuilding the entire system.

Finally, we’ll look at how success should be measured in this new architecture. Rather than focusing on simple deflection metrics, the conversation shifts toward self-serve completion, accuracy, and real resolution of complex queries.

In this practical one-hour session, we’ll cover:

How to assess LLM and RAG readiness before migrating your assistant
Identifying what should remain deterministic vs generative
Designing specialized agents and tool calls within a multi-agent architecture
Structuring knowledge sources across CMS, documentation, and APIs
Building modular systems that scale as new capabilities emerge
Measuring success through completion and resolution, not just deflection

If you're responsible for modernizing conversational AI systems, this session will give you a clear starting point for navigating the shift from legacy assistants to multi-agent AI.

Rachel is an experienced Conversational AI Consultant who specializes in customer-facing chat experiences for highly-regulated industries and government agencies. She is a business-savvy, human-centric designer who also has extensive experience with customer service, thanks to her early career as a licensed insurance agent. At CDI she enjoys building RAG solutions for clients and contributing to CDI Academy work through webinar development and hosting Unprompted episodes.

Olivia Marrese, PhD, is a Conversational Architect on the AI/ML team at 66degrees, where she engineers autonomous customer journeys and custom multi-agent ecosystems. By blending an academic background in linguistics and conversation analysis with deep technical expertise in agent orchestration, Olivia leads global brands across banking, travel, and retail in leveraging AI for high-impact customer experience.