<aside> ⚠️ Drift compounds silently. By the time users report problems, the system has often been degrading for weeks. Proactive monitoring catches it early.
</aside>
Objective: Detect when the AI system's performance, data, usage patterns, or alignment with objectives has shifted from baseline so the team can intervene before users notice.
When to use: Run on a regular cadence: weekly for high-risk systems, monthly for stable ones. Also run after any system update, data source change, or organizational restructuring.
Inputs: Baseline metrics from pilot, current performance dashboards, user feedback, system logs.
Position: Sentinel owns monitoring. Smith maintains dashboards. Architect investigates root causes when drift is detected.