SolarWinds’ Avior pairs AI agents with deterministic execution for enterprise IT

SolarWinds Worldwide LLC introduced Avior, an agentic artificial intelligence system designed to handle first-line operations for enterprise IT teams. The company aims to address the reliability issues that arise from AI agents producing different results upon repeated execution in production systems.

Chief Technology Officer Krishna Sai explained that Avior's design ensures "probabilistic models in deterministic execution," meaning the system's actions are grounded in a fixed framework of predefined goals and approval controls. These goals are based on 25 years of SolarWinds' experience with IT environments, and every workflow is fully auditable.

Avior operates as its own application on the SolarWinds Platform, reaching third-party systems through the Model Context Protocol. The system can merge related alerts into a single incident, identify configuration changes made outside normal channels and restore the last-known-good configuration on critical devices, and retire stale inventory records for devices no longer on the network.

SolarWinds emphasizes that each action in Avior runs with the permissions of the individual who initiated it and is logged. Rollback support is also built in. The AI system only sends data relevant to each request, and customer data is masked, never used to train the underlying models.

Company leaders believe that while reaction time alone is no longer sufficient in complex IT environments, Avior's tiered autonomy based on risk levels ensures that only explicitly approved high-risk actions proceed. Avior is accessible today through a guided onboarding program, though SolarWinds has not disclosed pricing.