Build Daily
Tinley Park · August 23, 2026Building Smarter AI with Context Graphs — Ben Roodman
Ben Roodman discusses the importance of context graphs in building smarter AI agents by capturing relationships, decisions, and outcomes that traditional databases miss. He highlights how Neo4j's community is helping enterprises implement multi-agent systems while keeping human relationships central to decision-making.
The outline
Tap a timestamp to jump there in the video.
Challenges with Traditional Data
Ben Roodman notes that understanding and describing relationships in CRM or transactional data remains a significant challenge for businesses.
The Role of Context Graphs
Context graphs provide a unified view of how people are connected, why decisions were made, and what the outcomes were, making them essential for next-generation AI.
Agent Memory and Schemas
The discussion explores what information, schemas, and ontologies agents need in their memory to make data actionable and understand human systems clearly.
Human-Centric AI Decisioning
Enterprises want to put agentic reasoning into practice while ensuring that human relationships remain a key part of the decision-making process.
Community Insights and Learning
Roodman reflects on the value of Neo4j's community in sharing insights about where systems are heading and expresses readiness to dive deeper into roundtable discussions.
Multi-Agent Systems in B2B
The video highlights how large organizations are thinking about specific B2B use cases for swarms of agents working together in enterprise environments.
Reviving Old Technology
Roodman concludes that context graph technology is 'old but new again,' providing the system people have been looking for to surface relationships and understanding in next-gen systems.
A context graph that captures why decisions were made and what outcomes followed gives agents more than raw rows from a CRM or transactional database. For a solo builder, this points to modeling the connections between people, choices, and results as first-class data if an agent is meant to reason about real work.