2019 was the year that AIOps graduated from skunk works shops to mainstream enterprises. At least that is what a recent, independent survey indicates.

Late last year, we paid a respected and independent third-party to survey over 270 senior IT executives and get the market's feedback on the state of AIOps. We produced a webinar in December that summarized the results and we are using data from that survey to inform this blog article. This is the first blog of a 2-blog series. Expect the second blog to be released in the coming weeks. 

AIOps Adoption into 2020

Not surprisingly, AIOps is following a predictable pattern, similar to Everett Rogers' classical model. What is surprising, is that it is at an accelerated pace when compared to the normal adoption cycle, we'll get into reasons for that later. So what did the data from the survey indicate? 2020 is the year AIOps moves from something reserved for 'early innovators' to a strategically viable technology for the majority. 

When looking at the figures above consider that a full 15% had already committed to implementing AIOps in 2020. But that isn't all that will implement AIOps in 2020, a portion of the 59% in the business case phase will as well - they just hadn't added it to their roadmap at the time of the survey. History suggests we can expect anywhere between 15% and 30% to move into the planning stage in 2020, so we could conservatively estimate 30% to 45% will be implementing by the end of 2020. 

What that means is that AIOps is real, and leading organizations are evaluating and implementing it in 2020. 

What that means is that AIOps is real, and leading organizations are evaluating and implementing it in 2020. 

Why Are so Many Organizations Considering AIOps?

So what has propelled so many into the planning and consideration phases for AIOps?  Across 2019 and into 2020 we see IT being squeezed. Many in the survey reported the same or smaller budget and staff size with increasing numbers of users and complexity. 

Some might argue that we in IT have dealt with this challenge before and yes, we have. What resulted was the adoption and ubiquity of virtual and then cloud services. We are also seeing increased popularity with the deployment of microservices and serverless computing, arguably for the very same reason.

Yet, the addition of these new technologies has added to the burden of complexity. And, the data suggests, IT organizations are already pursuing or will be starting some very complex IT projects in 2020, including: cloud migration, IT modernization, and digital transformation. In other words, the complexity is increasing exponentially.

The Challenges Outlined in the Data

IT departments are looking for ways to overcome challenges stemming from these pressures. 89% of respondents identified predicting and preventing IT Issues as a top IT challenge with 90% identifying correlating data across domains as a hot topic. 

Both of these response levels make sense, if organizations can predict and prevent IT issues then the need for their teams to scramble and resolve incidents is much less. Coupling prediction with correlation across domains helps IT teams spot issues and their true root cause well before the business is impacted. Both help to alleviate the staffing pressures and correlation helps to alleviate the growth in complexity. 

Can AIOps Address These Challenges?

The short but muddy answer is, "The traditional version of AIOps? Yes and no." The longer but less muddy answer:

Traditional AIOps tools act similar to the way event management tools operated many years ago. They aggregate, dedupe, and correlate events from a wide variety of eventing tools and sources, helping to find the signal in the noise. At the risk of oversimplifying, I'd say many AIOps solutions simply add AI on-top to help do those activities faster and with less human intervention (i.e. tuning). 

So, for many AIOps solutions, the answer is yes for "Correlating Across Domains." But, "Predict and Prevent" is where it gets muddy.

By relying on feeds from monitoring tools, traditional AIOps tools are relying on human-set thresholds to be crossed ahead of receiving an alert to correlate, dedupe, etc.  This leaves 2 possible holes:

1. Is everything that should be measured being measured? Consider all of the new technologies and services being used, are they all properly measured?
   
2. Monitoring tools are, by virtue of relying on thresholds, often reactive. They don't catch issues as they start to bubble up, but rather just before they are about to reach a business impact level. This makes working with events from monitoring tools to identify an issue well before business impact, difficult.

If traditional AIOps tools don't solve the main challenges outlined in by the IT professionals in the survey, how do we address those challenges?

A Solution Crying out from the Data?

The respondents didn't just have a grasp on their challenges, they also indicated a possible solution (to at least one of the issues). 50% of the respondents indicated logs as one of their most underused data sources.

Because logs show a change in behavior nearly instantaneously and give detailed information around the cause of that change, they might hold the keys to predicting and preventing incidents.  

And, just like any good multi-part Netflix show, I'm going to leave on you on that cliff hanger :). Don't worry, part 2 will come out soon and will discuss underused data sources a bit more, why those in the survey would consider looking at AIOps, what they expect to achieve, and what their concerns are around implementing an AIOps solution. 

More the binge-watching type? If you have 30-minutes you can watch the on-demand webinar. These and other findings were revealed in that presentation, not to mention the findings we will be discussing in part 2. 

Here's to a great 2020 for all of us in IT Operations!