It can definitely be clarified further. I would say that would happen no matter the data structure - monolithic graph, relational DB, etc.

The flow I had imagined (but not stated) was something like:

- Patient Record Agent compares Patient Record to Doctor/ Doctor Type that should treat this Patient.

- Having identified the Doctor/Doctor Type, it looks to seek doctor availabilities in the Doctor Availability Agent

- Having identified the right Doctor and the appointment availability, it then seeks the appropriate and available bed.

I had handwaved a number of actions that the agent must take to maintain the different graphs, which I think actually helps to illustrate the value of modular graphs more. Each agent would need to have a concrete set of actions to take on specific graphs/sub-graphs depending on what happened in the prior agent (i.e. doctor appointment booked), without having to worry about having cascading effects on a larger connected graph.