AIOps in the Cloud? Not So Fast

The promise of AIOps is to help enterprises detect IT problems more quickly by analyzing massive amounts of operational data faster than humans can and acting on the data to automatically fix the problems. This sounds great in theory. But CIOs aren’t buying it. Not just yet. The question is, why not?

To state the obvious, CIOs need to build and manage an IT infrastructure that is safe, sound, and secure. Their organizations demand it and keeping their jobs depends on it. This was tough enough when IT operations were exclusively on-site or in private clouds. It got tougher when organizations started migrating workloads to public clouds, creating hybrid-cloud environments that are more complex to manage.

Now, however, most organizations have moved beyond hybrid-cloud to multi-cloud environments that use public cloud services from more than one provider. Organizations evolve to multi-cloud for a variety of reasons, some by design and others by default. In multi-cloud by design, organizations choose to use different providers based on the specific characteristics of their business or workloads, or to meet data governance requirements such as those imposed by the EU’s GDPR (which mandates that customer data be held in specific locations), or simply to avoid being locked into one provider. Multi-cloud by default can happen when business units in an organization purchase a cloud service or a cloud-based application independently, without the knowledge of the IT department -- or multi-cloud can (at times, unwillingly) occur when IT is forced to pay for cloud costs by choosing a best-in-class application or service from a major cloud provider.

Regardless of how organizations end up with a multi-cloud environment, being there further raises the bar on the IT management challenge of unifying resources into a single, flexible, reliable, cost-and performance-optimized IT infrastructure. 

Where AIOps Comes In

AIOps (artificial intelligence operations) should be ideal in IT environments that are arguably too complex for humans to manage. Even though AIOps has improved by leaps and bounds over the past few years, CIOs aren’t ready to hand over their entire IT operation -- and put their jobs on the line if something goes wrong -- to a robot. Think of "War Games" or HAL in “2001: A Space Odyssey.” These movies depict our worst nightmares about what can happen when humans trust AI systems: hand over control, and then they malfunction.

Frankly, no CIO wants to be the first to have a major AIOps mishap and become a poster child for the dangers of trusting a machine’s recommendations over the judgment of human IT experts.

Still, there is little doubt that AIOps will become enterprise-ready sooner rather than later. The Big Four public cloud providers -- AWS, Azure, Google, and IBM -- are all in on AIOps. This is evidenced by recent developments such as IBM’s purchase of Turbonomic and the integration of AIOps into AWS. As Gartner noted, the long-term impact of AIOps on IT operations will be transformative. The firm predicts that by 2023, 30% of large enterprises will rely exclusively on AIOps and digital experience monitoring tools to monitor applications and infrastructure, up from 5% in 2018.

AIOps Today and Tomorrow

Today, AIOps adoption in multi-cloud environments typically starts with narrow use cases and focuses on observability, which is monitoring for and identifying problems and determining their root causes. This cautious approach lets CIOs evaluate AIOps for accuracy and reliability. As AIOps demonstrates proficiency, CIOs will gradually gain the confidence that is necessary to cross the chasm from observability into automated fixes of the problems that have been identified. The journey is in its early stages.

In the meantime, AIOps will continue to work alongside human teams of IT specialists, gaining experience and learning from their successes and failures in controlled environments as they augment their human counterparts while being overseen by them.

Jonathan (Jon) Cyr is the Vice President of Product Management, for Virtana, a SaaS platform providing AI-powered unified observability for migrating, optimizing, and monitoring hybrid and public cloud. He is a seasoned hybrid cloud product leader with a proven track record of delivering strategic cloud solutions across a variety of markets and customer needs to positively impact business revenues and operating efficiencies. Jonathan has 20+ years’ experience in architecting enterprise solutions for the IT Operations Management (ITOM) market.