The U.S. State Department is testing Agentic AI to accelerate cybersecurity operations and reduce administrative burdens, according to Ray Romano, deputy assistant director of the State Department’s Cyber Threat and Investigations division.

Speaking at a federal cyber resilience event, Romano said the agency is exploring how AI agents can help employees reclaim time by handling labor-intensive tasks. One key use case involves malware analysis, where an AI agent operating in a sandbox environment can reportedly analyze malware from a thumb drive in just 25 minutes, compared to nearly four days for a human analyst.

He described the initiative as part of the agency’s effort to prepare for a future increasingly shaped by autonomous AI systems. Unlike traditional generative AI, agentic AI can complete complex tasks with minimal human intervention.

However, he also noted that State officials are proceeding cautiously. The department is validating AI-generated outputs against human-generated work to measure both productivity gains and error rates before integrating the technology into day-to-day operations.

Furthermore, Romano also stressed the importance of governance, security, and organizational readiness, warning agencies against rushing toward new technologies without building the necessary guardrails.