Dynamics 365 Customer Service: Simulate AI field prediction with Case Management Agent

Field-prediction simulation – key capabilities

Administrators can run field‑prediction simulations using:

• Historical case records
• Uploaded Excel files
• Email or conversation transcripts

Simulation displays the Case Management Agent’s predicted field values based on the selected records or uploaded data.

Simulation using historical case records

Administrators select a set of historical cases to evaluate prediction quality. The condition builder allows filtering. For example, cases from the last seven days or cases related to Contoso Coffee Maker. After running the simulation, administrators can:

• View predicted field values.
• Compare existing field values with predicted values.
• Assess field prediction accuracy.

Simulation using email or chat transcripts

Administrators can upload emails or chat messages for prediction. Each row in the Excel file represents one email or chat conversation. The simulation outputs predicted field values for each row. Emails or chats can be extracted from the activity table and uploaded for bulk evaluation.

Simulation using Excel upload

Administrators can upload an Excel file where each row contains a single email or chat message. The system generates field predictions for each entry. Results help identify how the Case Management Agent would process incoming communications.

Evaluating and improving prediction quality

Simulation results help administrators validate prediction quality before enabling the feature in production. Based on results, administrators can adjust the Case Management Agent configuration. Simulations can be rerun after configuration updates to measure improvement.

This iterative approach helps identify optimization opportunities and ensures reliable predictions before the feature is turned on for the organization.