SolarWinds puts IT agents on a deterministic leash
Enterprise IT is where AI agents stop being a demo — a wrong guess there costs uptime, audit findings, and weekend callouts. SolarWinds' answer is to let the model decide what needs doing and never let it decide how far it can go.
SolarWinds today launched Avior, a standalone agentic AI system that takes on first-line IT operations work inside a fixed, auditable execution framework — the company's answer to the reliability problem of letting probabilistic models act on production systems. The model works out what needs doing; a predefined structure of goals, guardrails, and approval controls governs how it gets done, and every step is recorded. SolarWinds chief technology officer Krishna Sai described the design as grounding "probabilistic models in deterministic execution." Avior ships from today through a guided onboarding program on the SolarWinds Platform; the company did not disclose pricing.
At launch it handles a specific, unglamorous set of jobs: correlating related alerts into a single incident instead of a queue, detecting out-of-band configuration changes and restoring the last known-good config on the devices that matter most, and retiring ghost assets that no longer exist but still clutter dashboards and fire alerts. It also audits devices against a policy baseline with a record of every change, runs scheduled configuration backups, and tracks capacity and spend against the headroom a team wants to keep. SolarWinds says more capabilities are coming.
Autonomy is tiered by risk. Anything SolarWinds classes as potentially dangerous waits for explicit human approval, each action runs with the permissions of the user who started it, and rollback support is built in. Only data relevant to a request is sent, customer data is masked, and SolarWinds says it is never used to train the underlying models. Avior runs as its own application and reaches third-party systems through the Model Context Protocol, so teams do not have to migrate data to point it at self-hosted and cloud environments at once — the agents draw on 25 years of SolarWinds telemetry, topology, and configuration history.
The framing matters more than the feature list. Enterprise software has spent two years selling agentic AI on the model's judgment, while the buyers' actual blocker is repeatability: an agent that produces a different answer on Tuesday is a liability when an auditor asks what happened. SolarWinds is betting the product is the cage, not the model — and that bet cuts both ways. If the deterministic layer is the value, then ServiceNow, Datadog, and a wave of agent-governance startups are selling the same leash, and SolarWinds has to win on the operational knowledge it has already encoded, not on agent quality.
"The message from IT leadership is clear, reaction time alone is no longer enough," said Cullen Childress, SolarWinds' interim chief executive and chief product officer, framing the shift from managing alerts to orchestrating outcomes. SolarWinds' earlier assistants stayed inside individual products and, by the company's own description, were "bounded by human attention and prompting." Avior is the first one meant to run across the whole estate on its own.
What to watch: whether customers actually let it act without a human pressing approve. That threshold, not benchmark scores, is what decides whether agentic IT operations is a product or a demo. We covered adjacent ground earlier this year — Workiva ships no-code agents for work that has to survive an audit.
Would you give an agent standing permission to restore a config on your production network? Tell us in the comments.
