A compliance process can be automated, but not completely. The important thing is to separate three levels.
🟢 Fully automate:
- Task creation and routing
- SLAs, notifications, reminders
- Required-field validation
- Status control
- Version control
- Reviewer assignment
- Content review reminders
- KPI collection
- Audit trail
🟡 Automate with a human in the loop:
- Source verification
- Detection of potential ethical risks
- Risk classification
- Obsolescence checks
- Contradiction detection
- Preliminary editing
AI can assist the reviewer here, but it should not be the final judge.
🔴 Leave to people:
- Final ethical decisions
- Disputed interpretations
- Decisions about publishing sensitive material
- Material corrections with significant consequences
- Conflicts between commercial interests and customer interests
This is essential for trust.
Data Model for BPMS
CONTENT — content_id, title, type, status, owner, audience, customer_problem, product, risk_level, publication_date, review_date
CONTENT_CLAIM — claim_id, content_id, claim_text, source, source_date, verification_status, reviewer
ETHICAL_REVIEW — review_id, content_id, risk_category, decision, reviewer, rationale, date
CONTENT_VERSION — version_id, content_id, version, author, changes, date
CONTENT_METRIC — content_id, metric, value, period
CONTENT_FEEDBACK — content_id, customer, feedback_type, severity, resolution
This makes end-to-end quality analytics possible, from the brief through customer feedback.
💡 Final thought: BPMN here is not just a way to automate editorial work. It turns brand values into an operational product system: the value is defined at input, checked at every stage, measured after release, and fed back into the process. That is what distinguishes purposeful publishing from publishing for appearance’s sake.
Read the full series:
How Compliance Works: Ethics as a Process, Not a Slogan
How Compliance Works: Roles and Responsibilities
