Context Graphs: AI’s Next Big Idea

A growing conversation in enterprise AI circles centers on a concept called context graphs—an emerging idea that could define how intelligent agents operate within large organizations. The discussion originates with an essay from investor Jasmine Ball titled “Long Live Systems of Record,” which observed that as enterprise workflows become more automated, agents are only as reliable as the canonical data sources they pull from. The problem: in most real companies, canonical data does not exist in a clean, accessible form—ask sales, finance, and accounting for the same ARR figure and you will get three different answers.

Foundation Capital investors Jay Gupta and Ashish take this further, arguing that entire categories of enterprise knowledge are missing from any structured system—exception logic passed down through onboarding conversations, approval decisions made in hallway chats, and cross-system syntheses that live only in human memory. Their thesis is that a new class of infrastructure, the context graph, could capture these “decision traces” as agents execute workflows, building a queryable record of not just what happened but why.

The episode walks through detailed examples: a renewal agent that needs to know why a similar discount exception was approved last quarter, or how to handle an ARR figure that different departments define differently. For anyone building or deploying AI agents in enterprise environments, the context graph framing offers a useful lens for understanding why data governance and decision provenance will be as critical as model capability in the agentic era.

📺 Source: The AI Daily Brief: Artificial Intelligence News · Published January 06, 2026
🏷️ Format: Deep Dive