How Agentic AI Can Boost Cyber Defense

Transurban implemented an in-house agentic AI system—built on Anthropic's Claude and integrated with Splunk, ServiceNow, and AWS Bedrock—to automate SOC alert triage and ticket resolution verification. The design uses two agents (one for categorization, one for resolution verification) that provide full incident coverage while maintaining under 3% false positives; since deployment in September, triage times dropped 60% with 92% accuracy. Future plans include adding external threat intelligence and automating containment and response through Anthropic's Model Context Protocol (MCP).