What is AIOps and how does it work?

IT teams have their own version of TMI: Security operations center (SOC) teams frequently handle too much information, sometimes receiving thousands of alerts per day. IT operations may have to watch screen-fulls of help-desk tickets, application uptimes, and event logs.

The overload has led many tech departments to turn to two other letters—AI—to assist in combing through the info to provide quick insight and action. The use of artificial intelligence for IT operations is often referred to as “AIOps.”

“There is no doubt: There is no future of IT operations that does not include AIOps,” said a May 2022 report from market-intelligence firm Gartner, which predicted a $2.1 billion market size in 2025 and a compound annual growth rate of 19%.

What is AIOps?

AIOps platforms, as defined by Gartner, analyze telemetry and events, and identify meaningful patterns that provide insights and support proactive responses. 

The basic idea: Big-data tools comb through infrastructure, networks, applications, cloud environments, and existing monitoring tools to resolve IT issues, like an app going down or a spam message showing up.

Data and analytics tools from vendors like Elastic and Splunk allow operators to build their own AIOps-eration, companies such as DataDog and LogicMonitor augment their existing monitoring capabilities with artificial intelligence, and AIOps platforms from manufacturers like BigPanda and Moogsoft ingest events and respond to incidents.

How does it work?

IBM breaks the process down into three steps.

• An AI ops platform ingests data, creating a “normal” application baseline.
• Data is converted to action, based on past experiences, that can be sent to an IT pro or site reliability engineer via collaborative platforms like Slack.
• Finally: A script activates to resolve the issue.

For example…Say, you’ve got a mountain of help-desk tickets, specifically time-sensitive ones related to stock transactions. Anand Rao, principal with PwC's US advisory practice, helped an organization build an AI-ops model that spotted “easy-fix” tickets that could be remediated quickly.

“To be able to essentially sequence these resources in a way that was optimal, or at least better than what the humans were doing was a huge, huge gain for them,” Rao told IT Brew.

AI Ops drivers

New applications and services create new kinds of data, driving the AIops market.

“With data volumes reaching or exceeding gigabytes per minute across a dozen or more different domains, it is no longer possible, much less practical, for a human to analyze the data manually in service of operational expectations,” Gartner’s report said.

The most common AIOps frequently require the most real-time data, like financial services, media, and retail, according to Rao.

And AIOps is moving beyond its rule-based phase of finding trouble tickets and responding with preset actions.

Thanks to advances in generative AI, content-making bots can handle a lot more of the ops.

“It can create the test data for the code, then it can run the code in your Python interpreter or whichever environment that you have, and see if it runs or not,” Rao said. “So you're getting to the stage where the software is essentially going to adapt itself to whatever requirements that you're given.”