Kindo launches Deep Hat Sift to cut unnecessary AI agent runs

Kindo has launched Deep Hat Sift, a model that decides which incoming events warrant an AI agent run before the larger model processes them. Sift learns from the agent's verdicts, starts in shadow mode and activates only after a human approves a threshold, failing open if quality drifts.

• Sift filters events before the agent runs, cutting unnecessary investigations
• Starts in shadow mode; filtering requires human approval of a threshold
• Random sampling of withheld events measures ongoing miss rate
• Targets SIEM alerts, tickets, phishing reports and logs from Splunk, Datadog, New Relic