I started a new series on Medium: The Five Walls of a Context Layer.

If you’re giving AI agents access to your company’s knowledge, you’re building a context layer - the layer between your source systems and your agents. Build it yourself and you hit five walls:

  1. Knowledge graph & ontology
  2. Semantic layer
  3. Understanding intent
  4. Permissions
  5. Breathing data at scale

Part 1 is about the first and least visible one: your company has no map of its own things. The same service is billing-svc in Git, “Billing Service” in the wiki, BILL in Jira, and “the billing thing” in Slack - four names, one thing, and nothing that knows they’re one.

In the article I break down what an ontology and a knowledge graph actually are (the blueprint vs. the building), why built-in and custom entities are two different problems, and why an LLM wiki cannot hold your company’s knowledge once it grows past a single team.

👉 Read Part 1 on Medium

- Nave