Agentic GraphRAG

Chapter 8. Optimization

Your agent works. It reasons through multistep problems, queries live systems, and learns from its own mistakes. In a notebook, it looks like the future of enterprise AI. Then someone asks what it costs to run.

The invoice from a single week of production traffic, with every workflow node powered by a frontier model and every query routed through the most expensive option, is enough to shelve the project. But cost is only the first problem. An agent with unrestricted access to your knowledge graph can traverse from a public service catalog to internal cost data to employee records in a single query. A graph traversal that completes in milliseconds during development takes seconds at production scale, and a federation of specialist models that reasons beautifully offline becomes a latency bottleneck when an SRE is waiting for an incident diagnosis.

Production is where three forces collide: cost, governance, and performance. This chapter addresses all three.

We will start with selective intelligence, the principle of deploying the right model for the right task so that cost scales with complexity rather than uniformly. From there, we will cover the governance patterns that make graph-based agents auditable and secure: access control at the subgraph level, PII handling, and the execution graph as a compliance artifact. We will then turn to the operational concerns of keeping a knowledge graph healthy in production: schema evolution, lifecycle management, and ...