Your company already runs the hardest parts of an AI agent platform

The article proposes structuring enterprise AI agents like an org chart, with three inherited layers of company knowledge and policy.

Agents load only the documents needed for a specific action, reducing context size, cost, and inconsistency.

Human escalation and calibration happen through existing tools like Teams and enterprise directories.

The approach is meant to make AI agents easier to govern, cheaper to run, and more scalable across teams.