Conflict as Telemetry for Illegible AI is my peer-reviewed paper in the AAAI 2026 Spring Symposium proceedings. I delivered it as the keynote at the AI in Business symposium. The idea is that recurring conflict around a critical workflow (escalations, reversals, rework, blame cycles) is early-warning telemetry that the system has gone illegible. The paper lays out Map-Probe-Trace-Teach, the field protocol behind my Critical System Legibility Review.