It’s knowing how the enterprise works:
- how approvals actually happen versus what the process doc claims
- which exceptions are allowed
- who owns what
That dynamic is starting to shift with the rise of context graphs: decision-trace layers capturing who decided what, when, why, and based on what precedent.
Where LLMs commoditize functional expertise, context graphs democratize tribal knowledge and pull the “how shit gets done” from the shadows:
- the precedent that quietly became policy
- the real approval chain, not the org chart
- the “why” that dies in Slack threads
- the exception that became the rule
Key platforms are already moving in this direction. Microsoft, Salesforce, and Adobe are building native context capabilities. Glean, a startup focused entirely on this space, raised capital at a ~$7.2B valuation.
Context graphs won’t eliminate enterprise navigators. They’ll reprice them.
Practical starting point: pick one domain (approvals, exceptions, escalations) and make the trace queryable.
What tribal knowledge would you actually want to make machine-legible… and what would break (politically, legally, culturally) if you did?