AI System for Checking Patient Communication Permissions

The problem

In daily operations, permission records may conflict across systems. When channel permission, opt-out records, message purpose, preferences and record dates are spread across different screens, documents or team records, the conflict may not be visible in time.

This fragmented data delays preparation of the communication-permission check and forces staff to search again for supporting information. Before deciding, the data protection officer must complete missing fields and verify conflicts one by one.

A common exception is that the boundary between a service notification and a promotional message may require separate assessment. If this context is not checked, an incomplete or incorrect workflow could enable contact without permission or block a necessary notification, so its output cannot be applied directly.

What we set out to improve

• Number of records stopped because of conflicting permissions
• Rate of outputs corrected or rejected by the data protection officer
• Time between the source record and the review output

How the system works

1. 01 Define the review scope and the authorised data protection officer role
2. 02 Receive the channel permission, opt-out record, message purpose, preference and record date
3. 03 Check source identity, date, version and data freshness
4. 04 Label the message purpose
5. 05 Compare the most recent permission and opt-out records
6. 06 Check the exception that the boundary between a service notification and promotion may require separate assessment
7. 07 Place low-confidence or conflicting records in a separate human review queue
8. 08 Write the source identity, rule, output and time to the audit trail
9. 09 Have the data protection officer approve, correct, defer or reject the output

Methods we used

• A date-priority permission rule and purpose matrix
• Source, format, version and mandatory-field checks for channel permission, opt-out records, message purpose, preferences and record dates
• A missing-data warning instead of inference when inputs are absent or conflicting
• Separate assessment of exceptions and the main rule
• Joint retention of the source identity, rule result, communication-permission check and human decision

Where people stay involved

Data and security

Channel permissions, opt-out records, message purposes, preferences and record dates are accessed with the narrowest permissions required for the role. Identity and sensitive fields are masked where possible, and only the necessary section rather than the entire raw file is sent to the model. Access, output and the data protection officer's decision are logged; production use does not begin until the organisation approves data retention and location.

Who this suits

Frequently asked questions