Your AI agents are lost: give them a graph

The biggest limitation facing enterprise AI agents may not be the model. It may be the context surrounding it.

Anthony Alcaraz, Senior AI/ML Portfolio Growth Manager at AWS and co-author of O'Reilly's *Agentic GraphRAG*, joins Humans of Reliability to explain why reliable agents need more than a vector database and a large context window. They need structured knowledge they can navigate, memory they can prune, constraints they can follow, and feedback loops that help them improve.

Anthony breaks down how GraphRAG can support multi-step reasoning, preserve tribal knowledge, and give agents a clearer view of relationships across an organization. He also explores what happens when companies manage fleets of agents at scale, from smaller, more autonomous teams to roles that blend product, engineering, and customer work.

In this episode:

• Why understanding the business problem comes before choosing an AI architecture
• How poor context degrades model performance
• Where graphs add value to retrieval, memory, planning, and governance
• Why curated context can make smaller models more effective
• How to build self-improving agent systems with business feedback
• What agent fleets could mean for the future of teams and jobs

Find Anthony:

• LinkedIn: https://www.linkedin.com/in/anthony-alcaraz-b80763155/
• *Agentic GraphRAG*: https://www.oreilly.com/library/view/agentic-graphrag/9798341623163/

#AgenticAI #GraphRAG #ContextEngineering #ReliabilityEngineering