The high-level deployment architecture is provided in Figure 8.
A system logs model and a case level model are deployed in parallel. The case model is triggered when the actual case is created, and the system logs model is triggered after the logs are obtained and uploaded. The predictions produced by both models are exposed to CRM tool and an internal data platform. An intermediate Layer is also added where the predictions can be accessed in a data as a service model.
An instance of the CRM UI which contains the NBAs that the models produce and an option to provide feedback for the predicted recommendations is shown in Figure 9.