AIOps looms large as a way to help push the DevOps envelope. How can organizations prepare?
As organizations journey down the path of DevOps maturation, sustainable IT operations and IT service management remains a challenge for many. Even advanced organizations that have managed to speed up deployment rates and improve software quality struggle to maintain the resilience of the infrastructure that supports those applications. To support DevOps speeds, a growing number of organizations are turning to AIOps—the use of artificial intelligence (AI) and machine learning in IT ops—to speed up analysis of IT problems and better automate incident handling.
A new study out this week from OpsRamp shows that ops pros are able to reduce mean-time-to-resolution of incidents by as much as 50% through the use of AIOps. The use of AIOps is still infrequent, but analysts say its prevalence will grow quickly. Gartner says about 5% of organizations were using AIOps tools last year, but that by 2023 the deployment rate will shoot up to 23%.
But as DevOps teams lean on AIOps tools, management will need to harness a new set of skills to get the most out of the types of operational automation AIOps will afford. In the coming years, experts believe AIOps will change the complexion of the IT ops workforce. Fewer people will be needed in the network operations center (NOC) to manually sift through alerts and respond immediately to incidents. More people will be needed to curate data, scrub it and train the algorithms that will do the heavy lifting of event correlation and determining the root cause of problems.
As organizations and individuals plan out how they’re going to future-proof themselves for the changes in IT ops wrought by AIOps, experts say they should increasingly see the need for the following three roles in IT organizations.
AIOps Architects
Expect to see more job postings in the near future looking for AIOps architects and automation path designers, said Will Cappelli, CTO of Moogsoft and the former Gartner analyst who’s known for coining the term AIOps.
“(AIOps architect is) essentially a software development job involving the construction of meta-algorithms that guide the flow of control of the AIOps systems,” Cappelli said.
Potentially falling in this umbrella could be roles such as machine learning and AI designers that focus on designing the algorithms themselves, and interaction designers that focus on designing the automated workflows. According to Brendan Caulfield, co-founder of ServerCentral Turing Group, these roles already exist in other business capacities. With AIOps, these design jobs will start to be applied specifically to IT operations as well.
To prepare, veteran IT ops experts should start thinking about how they can bolster their skills, and management need to recruit and train accordingly.
“Mathematics and an end-to-end understanding of how modern IT systems behave will be critical,” Cappelli said.
ITOps Data Scientists
AI depends upon good, clean data and sound modeling to work effectively.
“The most sophisticated and expensive AI tools are useless if a company isn’t able to feed them with a high volume of high-quality data, in a format the system can digest,” said Rajesh Kalidindi, CEO of Leva Data.
This was corroborated by the OpsRamp report released this week that showed maintaining data accuracy is the No. 1 challenge cited by organizations deploying AIOps today. Many organizations will need ITOps specific data scientists to attend to this problem.
“Building high-volume, high-quality data sets will be increasingly important, and it requires a lot of human effort,” Kalidindi said. “AI has a long way to go before catching up with human reasoning (let alone surpassing it), and it can’t make any progress without people developing data models for AI to work with.”
Augmented Intelligence Analysts
According to Caulfield, algorithm training and support by domain experts in operations will become increasingly important at most organizations.
“No matter how ‘smart’ an algorithm is, or how well its designed, there will be exceptions that are unique to a business that will need oversight and tuning,” he said.
Kalidindi concurred that human intervention will be needed for a long time to augment and validate the insights and actions driven by AIOps, noting that this will lead to the rise of a new role for ops analysts.
“Especially in the early days of machine learning, humans are needed to judge whether an algorithm is leading to the right judgment,” he said. “For example, if we were to apply AI to analyzing the problems the IT Help Desk encounters, a human would be needed to judge whether the algorithm made an appropriate recommendation. If not, IT and data science needs to work with the human’s recommendation to better train the AI model on how to handle that problem in the future.”
One thing that most experts agree about is that the rise of AIOps will not diminish the need for IT pros. In fact, they say that domain expertise will be more important than ever. However, it will need to be bolstered by additional knowledge of data science, how AI works, and systems architecture.
“They need to retool themselves from admin-oriented activities to scripting/reporting-oriented tasks, and they need to be more application aware than ever before as infrastructure becomes a commodity,” said Ravi Raghavenderrao, vice president of service delivery at NetEnrich, of IT ops professionals in an AI-enabled future.
To those who have adjusted to DevOps transformations, this should all sound very familiar.
