The AI industry is celebrating ever-larger context windows — a million tokens, two, four — as though capability were simply a matter of wider doorways. Inside enterprises, the people actually shipping AI keep walking into a wall those doorways do not break.
The wall is not size. It is structure. Every company runs on two kinds of data: state — the closed deal, the approved invoice — and reasoning — the Friday-night call where the discount got approved, the precedent the support engineer remembered, the off-label note. State is what Salesforce, Workday, and ServiceNow were built to store. Reasoning lives in Slack threads, Zoom calls, and the heads of senior people most likely to leave.
A context graph is the missing layer that captures it. Foundation Capital calls it AI’s trillion-dollar opportunity. Neo4j’s Will Lyon calls it “the missing why.” I’d go further: it is an organizational world model — a structured, living representation of how a company actually thinks.
The full essay traces why incumbents structurally cannot build it, what it means for B2B product managers, and where the moats will form. Read on →
Previous Highlights
