Pratik Sarkar

Can Celonis Predict AI-Agent Failure Before It Happens? An Empirical Process-Mining Study of WildClawBench Execution Traces

This independent empirical study examines whether process mining can support early detection of AI-agent execution errors using Celonis and WildClawBench execution traces. The investigation covers 720 executions, 35,771 events, 12 AI models, and six task categories. The study dem...