A retailer scores every customer overnight for likelihood to churn and loads the results into a campaign tool the following morning. Separately, its website calls a model to decide within milliseconds whether a card payment looks fraudulent. Which description of these two workloads is correct?

• AThe churn scoring is batch inference and the fraud check is real time inference.
• BBoth are training workloads, since each one runs the model over customer data.
• CThe churn scoring is training and the fraud check is inference.
• DNeither is inference, because no new model is produced.