AI’s biggest promise for IT isn’t another faster answer. It’s helping teams move from directing every step to defining the outcome, setting the guardrails, and focusing their expertise where judgment matters. But an answer is rarely the end of an IT problem.
Someone must still understand what the answer means, decide whether action is appropriate, determine what should happen next, and verify whether the situation has improved. If AI creates another output that people must validate manually, it has added work rather than removed it.
That is why I see goal-oriented AI as more than another step in automation. It changes the relationship between people, technology, and operational work.
Goal-oriented AI lets people define the outcome they want, set the boundaries, and decide how much authority AI has. AI can then coordinate permitted work toward that outcome, evaluate progress, and return control to a person when judgment is required.
The leadership question is no longer only, “What can AI do?” It is also, “What outcome are we responsible for, and how should AI help us achieve it?”
The IT leader’s role is changing.
Most automation begins with instructions. A person identifies a problem, chooses a tool, starts a process, checks the result, and decides what happens next.
That model works when environments are small enough for people to hold the relevant context in their heads. It becomes harder as infrastructure, applications, dependencies, and operational data continue to grow.
Goal-oriented AI creates the possibility of a different operating model.
The IT professional still sets the destination. They still define the boundaries. They still decide what good looks like. But they do not necessarily need to direct every step between the starting point and the desired outcome.
That is the shift from operator to orchestrator.
An orchestrator does not step away from the work. The role becomes more deliberate. IT leaders and practitioners set objectives, establish policies, decide where autonomy is appropriate, review exceptions, and apply judgment when situations fall outside the expected path.
This makes technical expertise more important, not less.
As AI takes on more repeatable work, someone still needs to understand the environment, recognize risk, assess business impact, and know when an apparently simple action could have wider consequences. Those decisions require context and accountability.
The opportunity is to give experienced people more leverage over complex environments—not to remove them from the operating model.
Autonomy needs an operating model.
Autonomy is often presented as the destination. In mission-critical IT, I believe control and trust must come first. The right question is not whether AI should be autonomous, but where can AI can operate independently, where approval is required, and where human judgment must remain non-negotiable.
In IT, autonomy won’t be all-or-nothing. Teams need to decide which work AI can handle independently, which requires approval, and where people need to stay in control.
- AI can handle defined, lower-risk operational work independently.
- AI can investigate and recommend within broader workflows while people approve consequential actions.
- People can retain direct control over sensitive systems, unfamiliar situations, and changes with a wider potential impact.
The boundaries will not be identical for every organization. They may differ by environment, role, system, operational goal, or risk level. That is not a weakness. It is what makes autonomy usable in real IT environments.
For leaders, this means making several decisions deliberately:
- What is AI allowed to access?
- What actions can it take without approval?
- Which actions require a person to review the evidence first?
- When must AI explain its recommendation or record its activity?
- What should happen when confidence is low, or conditions change?
These are not questions to answer after deployment. They are part of the operating model.
Trust must be built into the system.
IT professionals are right to be cautious about AI. In a production environment, speed is valuable—but speed can also amplify the consequences of a poor decision.
For me, trust must be earned in practice: people need to understand what the system is doing, why it recommends action, and when it will stop and ask for help. It should operate within defined permissions, show the evidence behind a recommendation, record what it did, and verify whether the result improved. It should also know when to stop and ask for help.
We cannot ask people to trust a probabilistic system without transparency. Trust must be earned through the system’s design and its behavior in practice.
This is why governance cannot be treated as a checkbox or a policy document that sits outside the technology. Governance needs to shape the decisions an AI system can make, the tools it can use, the data it can access, and the points where a person remains in control.
At SolarWinds, this is the purpose of AI by Design. It establishes a foundation for privacy and security, accountability and fairness, transparency and trust, and simplicity and accessibility. The same principle guides our broader SolarWinds AI approach: AI should help teams work with greater clarity and confidence without creating unnecessary complexity or risk.
The standard for mission-critical IT should be straightforward: AI should augment IT professionals’ judgment, not replace it.
What should IT leaders decide before giving AI more autonomy?
Where can AI act independently?
Start with the work that is repeatable, measurable, and bounded. Investigation, correlation, evidence gathering, and recommendations may be appropriate starting points. More consequential changes may require explicit approval.
The important thing is to make the boundary intentional. “AI can take action” should never become “AI can take any action it wants.”
How will we know whether AI is helping?
Define success as an operational outcome, not simply a completed task.
A process can run successfully without improving service health. A ticket can close without resolving the underlying issue. An investigation can finish without producing a defensible answer.
Leaders should ask:
- What should improve?
- What must remain protected?
- What would show that the goal has been achieved?
- What would indicate that the system is moving in the wrong direction?
The more clearly success is defined, the easier it becomes for both people and AI to evaluate progress.
Where must a person remain in control?
Human intervention is not a failure of automation. In some situations, it is exactly how the system should work.
A production change, an unfamiliar condition, a sensitive system, or a decision with significant business impact may require human review. Approval and escalation points should be designed into the workflow from the beginning.
Does AI have enough context?
AI can only make useful operational decisions when it has the context required to understand the environment.
Applications depend on infrastructure. Infrastructure depends on networks. Changes can affect systems far beyond where they originate. If the relevant signals, relationships, history, and permissions are fragmented, AI will have a limited view of the situation.
That makes connected operational data and a strong observability strategy increasingly important. Context is not a supporting detail. It is part of the foundation for responsible autonomy.
AI should make IT more strategic.
The objective of goal-oriented AI is not to remove people from IT operations. It is to give them better leverage over the environments they are already responsible for.
As AI takes on more repeatable work, IT professionals can spend more time on decisions that require institutional knowledge, cross-functional context, and judgment. They can focus on resilience, risk, architecture, and the outcomes the business depends on.
This is not a future in which technical expertise becomes less valuable. It is a future in which the value of that expertise becomes more visible.
IT professionals will increasingly be the people who understand what AI can and cannot do, what data can be trusted, how autonomy should be bounded, and where decisions require human accountability.
That is why I see the move from operator to orchestrator as a leadership transition—not a reduction in IT’s role.
Where SolarWinds® Avior™ fits.
At SolarWinds Day, we introduced the idea of AI as role-based teammates. Avior applies goal-oriented AI to IT operations, with teams defining the outcome and the guardrails.
Avior is goal-oriented AI for IT operations. It is designed around the model described here: teams define an operational outcome and the guardrails, while role-based AI agents gather evidence, coordinate work, take permitted actions within the user’s permissions, verify results, and hand work back when human judgment is required.
Avior is one example of how AI can help teams move from managing individual tasks toward directing outcomes—while keeping people in control of the goal, the boundaries, and the decisions that matter.
See what goal-oriented AI looks like in practice: Learn More.
