Elastic on Thursday, Oct. 8, announced a private preview of Elastic nightshift AI SRE, an AI agent built into Elastic Observability that watches telemetry around the clock, investigates incidents and shows the reasoning behind its conclusions, according to the company's press release.

The pitch is aimed at problems that never set off an alert. Elastic cited an April 2026 NeuBird AI study finding that 78% of organizations had at least one incident in the past year where no alert fired and customers found the problem first.

Elastic’s Nightshift AI SRE Agent Puts Transparent Reasoning at the Heart of Incident Response

Four engines

Nightshift is built from four engines that can work together or separately: Context, Detection, Investigation and Remediation, Elastic's Luca Wintergerst wrote on the Observability Labs blog. Teams can, for example, have it investigate alerts from the monitoring tools they already use without turning on detection.

The Context engine builds knowledge of an environment from telemetry, code, runbooks and wikis, and keeps a long-term memory of past incidents. Detection evaluates telemetry continuously, with no alert rules to write, and groups related signals into a "significant event." The Investigation engine can start from any alert, in Elastic or a connected tool. It forms hypotheses, tests them in parallel against logs, metrics, traces and profiles, drops the ones the evidence doesn't support and reports a root cause in plain language, along with affected services, proposed actions and the blind spots it couldn't verify.