context graph

Gate C8, flight 34/185, four seats left, nine minutes before the door closes: that is the slice Rakesh Gohel calls a context graph. Gohel, who builds governance infrastructure for AI agents at JUTEQ, separates it from the knowledge graph, which holds what is true at company scale (passengers, aircraft, flights, fares, routes, airports, weather, policies) and is kept stable and governed. The context graph extracts only what the decision in progress needs, including how similar cases were handled before. His instruction is to scope the graph at each step rather than retrieve everything.

Jaya Gupta and Ashu Garg of Foundation Capital arrive at the same object from the enterprise software side. They locate a missing layer beneath AI agents: decision traces, the exceptions, overrides, precedents and cross-system context that currently live in Slack, deal desk conversations, escalation calls and employee memory. Accumulated, those traces form a context graph, a living record of decisions linked across entities and over time, where precedent becomes queryable. They argue the incumbents cannot build it. Salesforce stores current state, not state at the moment of decision. Snowflake receives data by ETL after decisions are taken, losing the decisional context.

The two notes converge on a relocation of the system of record, from the business object to the decision. Gohel's infographic places the context graph third in a five-link "Semantic Spine", between the knowledge graph and the procedural graph. Gupta and Garg leave open which startups end up owning that layer.