Building Multi-Agent Systems In Llama Index

In this forward-thinking session from the Artificial Unintelligence Conference 2025, Laurie Voss (LlamaIndex) explores the next frontier of artificial intelligence: multi-agentic systems. Moving beyond monolithic models, Laurie shares how networks of smaller, specialized AI agents can collaborate to solve complex problems more efficiently, accurately, and transparently.

Drawing on insights from thousands of real-world implementations built with LlamaIndex, Laurie reveals what makes multi-agent systems work — and what can make them fail. The talk blends technical rigor with practical wisdom, showing how communication, coordination, and human oversight form the foundation of scalable AI ecosystems.

Key themes include:

Why the future of AI is multi-agent, not monolithic

How specialized agents improve efficiency and precision

Designing effective communication and context-sharing protocols

The ongoing importance of human oversight and governance

Lessons learned from thousands of LlamaIndex-powered projects