Your Agent's Problem Isn't the Model - It's the Knowledge Layer

Most enterprise agent projects bolt an LLM onto a data lake, let it query raw sources live, and then act surprised when the thing is slow, wrong, and impossible to audit. The model was never the bottleneck. The knowledge layer is — the engineered substrate that stores, updates, retrieves, and composes an organization's internal knowledge for the agent. It's the part that decides whether the system is reliable in production, and it's a discipline in its own right, separate from the prompt and separate from the model. This talk maps that layer end to end. I'll walk the four sub-problems — storage, updating, retrieval, context construction — and the techniques teams actually reach for against each: advanced RAG, GraphRAG, agentic retrieval, and agent memory. Each gets mapped to the mainstream products you'd build it with, grounded in patterns from real production systems. You'll leave with a vendor-neutral way to decide which layer your problem actually needs — and which expensive upgrade you can skip.
Format: TECHNICAL TALK, BUSINESS TALK
Location: ROOM B
Speakers: Ratko Nikolic
Level: Beginner to Intermediate
Audience: Registered participants