A system and a method are disclosed for receiving an entry, the entry comprising first content and a metadata tag corresponding to a classification, the first content populated by a first source. A rules engine determines that the first content comprises a data field associated with at least one of a plurality of reconciliation policies. Responsive to determining that the first content comprises a data field associated with at least one of the plurality of reconciliation policies, the rules engine selects a reconciliation policy based on the metadata tag. The rules engine retrieves, from a second source, second content associated with the data field, inputs the first content and the second content into a model, the model selected based on the reconciliation policy, the model generating an output, and performs a remediation action based on the output.