About the session
As enterprises move beyond GenAI experimentation, the real challenge is no longer "what models to use," but how to build systems where AI can reliably operate on enterprise data, make decisions, and remain auditable.
This session presents a practitioner's perspective on designing and deploying Agentic AI systems grounded in enterprise data platforms. Drawing from real-world implementations, it will walk through how foundational data systems (structured, semi-structured, and unstructured) can be made AI-ready, and how multi-agent architectures can be layered on top to enable intelligent decision-making workflows.
The session will cover:
• Designing AI-ready data platforms that support retrieval, reasoning, and traceability
• Building multi-agent systems (classification, retrieval, critique, and orchestration layers) using frameworks like LangChain and LangGraph
• Applying RAG + critique loops + routing architectures for enterprise use cases such as compliance and policy enforcement
• Ensuring auditability and governance through traceable pipelines and custom observability/visualization layers
• Key lessons from production environments (including failure modes, guardrails, and scaling considerations)
Rather than a theoretical overview, this will be a deep dive into how Agentic AI systems are actually engineered, deployed, and governed in enterprise settings, with architecture patterns and walkthroughs that attendees can adapt to their own organizations.
