accessing a training set of information describing entity characteristics of each account holder entity of a first set of account holder entities, content item characteristics of a content item presented to each account holder entity of the first set, and interactions between the first set of account holder entities and the presented content item, each account holder entity associated with one or more central database system accounts;

training a machine-learned model using the accessed training set of information, the machine-learned model configured to identify additional account holder entities that, if presented with the content item, are most likely to interact with the content item;

identifying a target set of account holder entities, the central database system storing information describing entity characteristics for each account holder entity of the target set of account holder entities in a database;

applying the trained machine-learned model to the entity characteristics of each account holder entity of the target set of account holder entities, the entity characteristics of each account holder entity of the first set of account holder entities that previously interacted with the content item, and the content item characteristics to identify a subset of the target set of account holder entities for presentation of the content item;

causing display of the content item to the subset within an interface displayed by a device, each device associated with an account holder entity of the subset of account holder entities, the content item including an interface element that, when selected, causes an interaction to take place;

receiving one or more interactions from an account holder entity with the displayed content item;

updating the training set of information to additionally include the received one or more interactions and associated labels indicating for each of the received interactions that the interaction is a positive interaction or a negative interaction; and

updating the machine-learned model by retraining the machine-learned model using the updated training set of information.