The Context Graph Panel with Will Lyon, Jessica Talisman, Dave Bennett, Yann Bilien and Emil Eifrem

The conversation kicked off with a simple but important idea: before we go too far, let’s make sure we’re actually using the same words to mean the same things. Terms like “context” and “knowledge graph” get thrown around a lot, and they can mean very different things depending on who you ask. One common thread from the panel was that a knowledge graph is the broader foundation — your structured domain data like people, organizations, events — while a context graph builds on top of that by adding things like decision traces, workflows, and institutional know-how. In other words, it’s not just about what’s true, but how decisions get made. A big reason this is all resurfacing now is the rise of LLMs, which make it much easier to automate parts of the setup and maintenance that used to be painfully manual. But the group was clear: the tech alone isn’t the hard part. The real challenge is doing the knowledge management work — modeling your data well, resolving entities properly, and actually capturing how humans make decisions — so agents can operate reliably in the real world.