An agentic enterprise is an organization that uses autonomous AI agents to execute and coordinate business processes. Instead of a single chatbot answering prompts, multiple agents may plan tasks, call tools, move data between systems, and escalate work to higher-level supervisors. This can improve speed and scale, but it also creates a new operational attack surface because the agents can act with real permissions.
In cyber security, the main risks are prompt injection, tool abuse, excessive privileges, data leakage, and runaway agent loops that trigger repeated actions or expensive API calls. Defenders use least privilege, approval gates for high-impact actions, logging, anomaly detection, and strict tool allowlists to reduce harm. In practice, an agentic enterprise needs the same controls as any other production system, plus continuous monitoring of what the agents are doing, why they are doing it, and whether their actions are safe and authorized.