eSentire compresses expert threat analysis from five hours to seven minutes

eSentire, a managed cybersecurity firm, built a threat-investigation agent on Claude via the API that reasons over its own alert and telemetry data. Vendor-reported outcomes: expert analysis time drops from about five hours to seven minutes, with the agent agreeing with senior human analysts around 95% of the time. In security operations the build-on-API path wins when the value is domain-specific reasoning over proprietary data rather than a generic support flow, and the human expert stays in the loop as the check on the remaining 5%.