AI-assisted grade of how each level handled every closed escalation, against the L2-to-L3 guideline. This is a review aid, not an official performance score. Every row shows a confidence flag; you can regrade or override.