Cybersecurity provider Sophos announced it has reduced its average threat investigation time from 38 minutes to 89 seconds by deploying custom AI agents built through OpenAI Daybreak. Integrated into its Sophos Fusion defense platform, the multi-agent system automatically synthesizes sensor data across 500 integrations, formulates investigation plans, and proposes remediation steps for security operations teams.

According to Sophos, the agentic workflow currently automates roughly half of all incoming cases, distilling trillions of daily events into actionable investigations across its nine global security centers. The platform offers customers three operational modes ranging from advisory notifications to direct automated remediation.

To balance rapid response with risk management, Sophos maintains human oversight for sensitive operations, requiring security analyst approval before executing destructive or high-risk defensive actions.

Why it matters

  • Demonstrates concrete enterprise ROI for multi-agent LLM systems in high-volume, time-sensitive operational environments.

  • Provides a template for human-in-the-loop control models in mission-critical automated workflows.

Source: openai.com