Enterprise artificial intelligence (AI) risk has materialized in ways that traditional audit playbooks were not built for. Historically, AI followed a narrow loop wherein a user submits a prompt and the model provides an output. This is where the exchange ended. Agentic AI, however, has altered that loop. Large language models (LLMs) now plan and execute multistep workflows, read and write files, and update records through autonomous decision chains, all with little or no human review....