The Evolution of AI in IT Ops: Exploring the Promise of AIOps
A private executive event happening this Thursday has me thinking about the real state of AI in ITOps and whether AIOps is the real deal. Or not.
For IT Ops leaders, there’s a certain irony in the fact that artificial intelligence (AI) is now the hot topic in business circles.
Of course, AIOps has been a thing since long before this recent AI craze.
Although, there are plenty of skeptics who will say that AIOps wasn’t really about AI. After all (as I understand it), the first iteration of AIOps didn’t exactly stand for artificial intelligence in Ops. It stood for algorithmic infrastructure and operations.
So, which is it? Have all these AIOps vendors just been posing this whole time, waiting for the real deal to show up and show them up? Or was IT Ops the OG AI playa?
The answer, of course, is yes!
AIOps was — and is — the real AI deal. But the new ChatGPT-powered AI craze also opens the door for IT Ops leaders to re-assess and reimagine their use of AI within the domain of IT Ops.
Exploring this re-imagining is the topic of a private, invitation-only executive event I’m hosting this week, but I felt it was worth a little pre-gaming here.
Is the AI in AIOps Really AI?
Part of the reason that we’re even having this discussion right now is two-fold.
First, artificial intelligence may be one of the most nebulous terms ever created. The press (and even us pundits that should know better) throw the term around as if it is a single piece of technology (I’d like three orders of AI, please!).
At best, it is a term that represents a vast diversity of technologies that lack any real agreed upon definition (not that there isn’t a definition, but that there are hundreds of them).
Most industry observers tend to agree that AI signifies a process by which a machine (e.g., a piece of software) is interacting independently in some fashion.
Man, even that last sentence was difficult to write because I am literally bracing for people to challenge it — despite its incredible broadness!
The reality is that AI is incredibly difficult to define. For me, the simplest way to think of it is to contrast it with automation. Automation is a machine performing a pre-determined action that I have specifically programmed it to do. AI is when a machine is delivering an outcome that I have prescribed it to deliver, but which specific actions to deliver that outcome I have not.
I know, that’s still a slippery slope, but it’s the best I can do at the moment. (Editor’s note: If you have a better definition that you’re willing to defend 😀, I’d love to hear it! Please post it in the comments.)
Using this definition, I believe that AIOps does, in fact, leverage AI.
Yes, most AIOps tools are predominately using algorithms and more basic forms of machine learning (another loaded term) to function, but the net result is that AIOps tools focus on delivering specific, functional outcomes (e.g., the reduction of alert storms) by collecting data, analyzing it, and then determining the best course of action to deliver the outcome.
There is no simple script that says filter all alerts that look a certain way. This ability to dynamically analyze and respond to changing data sets or environmental elements is what separates something like AIOps from something like Robotic Process Automation (RPA), which relies heavily on fragile, pre-determined scripts.
The Evolution of AI in Ops
Still, many AIOps tools have struggled to move beyond performing the simpler, yet heavy-lifting tasks such as managing alert storms. The movement towards leveraging various forms of AI to enable greater autonomous automation and self-healing has been a much slower slog.
There are a few reasons for this slower adoption of broader AI use cases.
The first is IT Ops leaders and operators themselves. Many have simply been skeptical of the veracity of these sorts of tools and lacked the trust to put them into production in critical use cases without direct human intervention and oversight.
It’s a fair concern, and one that many vendors themselves advocated. However, the culprit has mostly been a combination of:
The lack of reliable operational data;
The unwillingness to create operational training data sets to tune AI models based on a company’s specific operational environment;
And — most importantly — a fear that the promise of improved efficiency was outweighed by the political risks of a catastrophic failure while the IT operational machine was left to run on autopilot.
While the first two issues remain and need to be addressed, it is the third that was most significant for many IT leaders — and the one that the sudden rush of AI mania is most likely to help overcome.
The rise of ChatGPT has, as they say, captured the imagination of business leaders everywhere. In some cases, you will be challenged as to why you’re behind the AI curve. And in others, you may have to explain why a blanket “ban on all AI” is not only misguided, but non-sensical (given our earlier discussion on the lack of any clear definitions).
But the point is that it is no longer an obscure conversation. Whether you’re talking about AIOps or almost any other use of AI in the IT Ops estate, ears will perk and people will be paying attention.
So, what should you do with this newfound interest?
Critical Next Steps for the IT Leader
As an IT leader, this is a time to step up and stake a claim.
This is NOT the time to pfft, pfft and try to explain that all this AI talk is just hype!
AI is a topic that will continue to both fascinate and confound for the foreseeable future. That fact will provide ample opportunity for you to take on the role of educator and guide without having to force the issue.
Secondly, and even more critically, getting AI right is going to be a critical driver of competitive differentiation for every enterprise. That’s obviously going to be true in a vast array of customer facing uses, but it will be just as important within the function of IT.
AIOps has long promised things like self-healing, improved efficiency, and reduced workloads, yet has struggled to fully deliver. Its rapid evolution, particularly when combined with the adoption of other approaches to AI, offers a pathway to fully realizing its promise.
At the same time, the heightened awareness of the risks of AI may also prove to be a boon, giving you the opportunity to rationally mediate the application of this new class of technology for the good of the enterprise.
The fact that IT Ops has been at the forefront of the adoption of AI within IT puts IT leaders at the forefront of this vital conversation.
Still, we’re at the beginning of this process, not the end — which is why I expect our conversation this Thursday will be lively and engaging. Want to join us?