The model is the same everywhere. What changes is the entity it learns, the events it reads and the decisions it returns.
Decisions at transaction speed, taken on the raw event history of each account instead of on aggregates that were already stale when they were computed.
Every interaction is a signal. Pricing, recommendation and demand read the same event history, so they stop contradicting each other.
Network and customer data in one model, so the operation can act before a problem reaches either the customer or the report.