Part 8 of this blog series offered expert predictions on the future of AIOps, and automated remediation was one of those future expectations. To delve deeper, APMdigest asked the experts: Is auto-remediation the ultimate goal of AIOps, and is this practical or even possible?

Start with: Discovering AIOps - Part 1

Start with: Discovering AIOps - Part 2: Must-Have Capabilities

Start with: Discovering AIOps - Part 3: The Users

Start with: Discovering AIOps - Part 4: Advantages

Start with: Discovering AIOps - Part 5: More Advantages

Start with: Discovering AIOps - Part 6: Challenges

Start with: Discovering AIOps - Part 7: The Current State of AIOps

Start with: Discovering AIOps - Part 8: The Future of AIOps

Human Intervention

While some experts foresee a future where AIOps will be able to automatically remediate issues, today the focus is on providing humans with the information to take action themselves.

As of now, the enterprises are divided in a few different stages when it comes to AIOps adoption, according to Monika Bhave, Product Manager at Digitate.

First is the manual category, where there's complete human dependency to perform tasks, with high risk and minimum efficiency. Next is the assisted category, which is defined by machine-assisted tasks that require lesser human input. This phase is suitable for environments where a set of similar tasks need to be performed over and over. Many enterprises are still at the manual or assisted category.

"Longer-term there is of course the notion that advanced AIOps might and should include systems that are somehow responsible for practical handling of all of this in a coordinated fashion — something like self-healing where the systems themselves can identify and fix issues as they emerge, but, as with most areas of automation today, the work now is primarily focused on providing the right information and context to the right human user, at the right time, to improve and accelerate existing workflows," explains Asaf Yigal, CTO of Logz.io.

While AIOps can automate routine tasks and provide valuable insights, certain complex decisions and strategic planning still require human expertise, adds Bharani Kumar Kulasekaran, Product Manager at ManageEngine. For example, while AIOps can automate certain repetitive manual tasks, more critical activities, such as pushing a configuration change, is something that needs to be closely monitored by an actual IT admin.

"The vast majority of the time, when something has gone terribly wrong, you will need smart people to leverage experience to keep things running," asserts Heath Newburn, Distinguished Field Engineer at PagerDuty.

Automated Remediation Today

"Auto-remediation happens today but it's mostly for low-impact, repeatable things. You don't even necessarily need AI for that," says Shamus McGillicuddy, VP of Research, Network Infrastructure and Operations, at Enterprise Management Associates (EMA).

Most enterprises are looking to use automated remediation on tasks that are repeatable and well defined, Brian Emerson, VP & GM, IT Operations Management at ServiceNow, agrees. AIOps deployed today can already remediate some tedious or simple tasks without human intervention, either through enhanced self-service capabilities or through actual AI-driven remediation.

"I'm aware of many enterprises and MSPs that have been doing automated remediation for six months and a couple for over a one year. They're not doing it for everything in IT but they've identified and tested specific use cases where they have fully autonomous actions from detection to remediation in place," affirms Carlos Casanova, Principal Analyst at Forrester Research.

"AIOps is all about improving accuracy while optimizing human expertise at this point. Perhaps at some point it really is about eliminating the need for human expertise, but I don't think we're very close to that yet as the systems are still only as good as their users," says Yigal from Logz.io.

A Long Way to Go

"AIOps solutions could potentially enable autonomous, self-healing IT infrastructure, but we're probably 6 to 10 years away from that being a reality," Shamus McGillicuddy from EMA predicts.

"In terms of creating AI-driven systems that understand how to monitor and troubleshoot themselves in an automated manner, there's likely still a long way to go until anyone is willing to trust the system to do so — if that is ever really going to make sense from the standpoint of giving up more control to these AI capabilities," Yigal says.

"But I do think we will start to see systems that are trusted to do some low level decision making in their own right, and this is essentially occurring when we allow the system to decide what is good information to rule out from a troubleshooting perspective, for example," Yigal from Logz.io continues.

"One of the key trends for us to consider in the world of observability is observability-driven design, where the developers are building their systems with the direct purpose of making it optimal for observability purposes. That's a movement that will certainly help platforms work smarter and potentially get to automated remediation, etc. based on the increased level of understanding and precision; but we're not there quite yet," Yigal adds.

Looking to a NoOps Future

"There's been a level of skepticism over the last 15 years regarding automation. Much of this skepticism is a result of underwhelming results from early AIOPS entrants that focused their intelligence (correlation etc.) on strictly alert data," Payal Kindiger, Senior Director of Product Marketing at Riverbed, recounts.

"However, with the combination of Unified Observability platforms that utilize AI/ML techniques, we have the ability to counter this alert-driven intelligence approach by applying runbook automation to full-stack, full fidelity telemetry data. With these combined capabilities, the potential for automation mimicking intelligence and expert decision making and logic while ingesting actionable insights across the IT ecosystem is much broader than it has ever been before."

Bhave from Digitate says, "I use the term autonomous enterprise. Within the next 5 years, AIOps will provide the foundation for a fully autonomous enterprise. Here, AI and machine learning detect and resolve all IT issues automatically — and do so without IT even knowing something's gone wrong and is being fixed. It's also important to note that with the predictive maintenance capabilities of AIOps, there will be fewer issues that need to be resolved in the first place."

The term "NoOps" is often used in this context, meaning automated IT Operations that does not require human intervention. "Automated ops, or NoOps, is definitely on the horizon," Yigal from Loz.io foresees. "In fact, with the rising volume of AI-generated threats and code, we absolutely need to be talking about NoOps because we're headed toward a future in which humans simply won't be able to handle the volume."

Saying No to NoOps

Several of the experts disagree with the NoOps vision, however.

"The notion of a NoOps future is far-fetched at best," cautions Dennis Drogseth, VP at Enterprise Management Associates (EMA). "Instead, what we see is the need for less siloed ways of working, innovation in leveraging analytics and automation to improve existing processes, and increased awareness of the business-to-technology handshake."

"No, not now, and likely not ever. We do not have ML models powerful enough to do this with any degree of reliability for real-world applications. We will always require human oversight and a human-first AI approach," says Phillip Carter, Principal Product Manager at Honeycomb.

"No matter how sophisticated AIOps gets or widely it gets adopted, I don't see a scenario where NoOps is a wide-spread reality," Carlos Casanova from Forrester agrees. "Will there be areas within an enterprise where AIOps runs fully autonomously? Sure, but in small controlled settings."

Automated remediation can handle predefined scenarios, but the ever-changing landscape of IT operations demands the human element for critical decision-making, innovation, and adapting to unforeseen challenges. Human-aided AIOps, where AIOps augments human capabilities, is more likely than a fully autonomous NoOps future, says Kulasekaran from ManageEngine.

"There will be continuous improvements over time, but we feel the value is helping the IT teams do their jobs better and more effectively rather than assuming human workers will be eliminated from the operations process," says Emerson from ServiceNow.

"I don't think we'll ever see a NoOps future. There will always be a human touchpoint, but with AIOps we can get much better at managing the complexity and chaos," Bill Lobig, VP Product Management of Automation at IBM, concludes.

Making AIOps People-Centric

"We don't talk enough about people in AIOps. This is one of the reasons that people are so wary of it. AI is not going to save us," Newburn from PagerDuty admonishes.

"Autonomic computing hasn't happened yet," Newburn continues. "Even the most highly automated organizations are leveraging smart people to fix problems well more than half the time. We need to refocus AIOps as people-centric, meaning arming people with better context, decision making, and guided automation from wherever they want to work. When we do that, AIOps can really achieve its promise."

Go to: Discovering AIOps - Part 10, the final installment in the series, with tips on getting started and succeeding with AIOps.