2026 in AI Ops presents opportunity, challenges

It’s a new world for AIOps as the use of artificial intelligence in operational technology accelerates.

In 2026, IT pros should expect to see AI usage grow in importance, said AppOmni Director of Artificial Intelligence Melissa Ruzzi, as automation and “self-healing” become mainstream. IT teams seeking to leverage AI into operations are in a good spot as technological advances become increasingly practical.

“The combination of all the other AI tools, traditional machine learning that I see that are already used on AIOps, with the generative AI on top of it, making decisions,” Ruzzi said.

Full scope. Planning for 2026, NCC Group Technical Director David Brauchler told IT Brew, is important—he stressed that AI has been used in IT operations for over a decade. What’s new is how powerful the technology has become, especially for defending operations from attackers.

“We have these powerful new paradigms of AI that we’re beginning to ask interesting questions about what it means to trust an AI system to operate our defenses on our behalf,” Brauchler said.

That process includes allowing AI systems to run incident response playbooks on behalf of the operator, but it’s a delicate dance. An attack on an AI system itself could compromise operational capabilities and undo the benefits of the technology.

“I do predict that we’ll see businesses begin to give AI systems more functionality, more tools, and, frankly, more power to respond in the same way we would expect a sysadmin or one of our security operation center team members to respond,” Brauchler said. “If AI is given all of these tools, the big question is, how are we preventing these AI components from themselves being manipulated by the threat actors interacting with our systems?”

Lock down. As Brauchler notes, it’s a critical question as AI continues to have a role to play in the IT operations space. The more information and decisions that are turned over to autonomous AI agents and systems, the more the attack surface expands. It’s a matter of managing the balance correctly.

Ruzzi noted that security is an ongoing concern, especially when giving agents the level of autonomy they need to be self-healing. If AI agents are capable of taking actions up to and including restarting servers, for example, then there’s a possibility that those permissions can be turned against users.

That’s why human control will remain important, and help AI systems to improve.

“Whenever the AI is not sure what to do, that’s when the expert jumps in,” Ruzzi said. “Or if AI made a decision and was not the best decision, the human can jump in to help AI to learn what it should have done—and then the next time it’s going to do better.”

Cold water. Not everyone, however, is as bullish on the role of AIOps going forward. Richard Bird, Singulr AI CSO, told IT Brew that he thinks AIOps could take an outsize role in the 2026 IT landscape. That’s despite a lack of clarity whether AI investments are actually making a positive difference in operations.

“Going into next year, understanding whether or not you’re getting a return on investment is going to probably be one of the top three problems of 2026 when it relates to AI,” Bird said. “There really has been no meaningful progress made on figuring out how to measure it.”

Bird is also skeptical that AIOps has a lot to offer—or, at least, that we know which opportunities exist. He noted that blockchain was formerly promised as the be-all and end-all solution, but turned out to be expensive and onerous to integrate into existing systems.

“That’s going to be the continuous grind next year, which is still a whole lot of on-the-job learning, a whole lot of effort to quantify value,” Bird said.

Asking questions about what works is part of determining the best use cases for AIOps. Bird said that skepticism is warranted, and that for all its promise, AI has yet to really deliver on the operations side.

“We all need to continue to be skeptical of technology,” Bird said. “When we’re presented with a story about AI that says it’s easy to implement, just put it in place and it’ll take care of everything—I haven’t seen any AI that’s worked that way yet.”