Content Copyright © 2022 Bloor. All Rights Reserved.
Also posted on: Bloor blogs
Last week we hosted a small roundtable discussion over dinner in London with some IT leaders to get their views on AIOps…what they think it is and how well they view it. We did this in conjunction with AIOps vendor, Big Panda, who wanted to understand customer sentiment in the market as they look to grow their presence in Europe.
There was general consensus on the challenges faced by IT operations teams in dealing with the increased complexity of modern IT infrastructure and providing assurance to the business for its performance and availability. There was no argument about the need to be able to handle the flood of alerts from multiple different monitoring and management tools effectively or being able to correlate data from multiple sources quickly and accurately to speed up root cause analysis and reduce mean-time to resolution (MTTR) (Have a look at our research and blogs on the issues for a deeper dive).
Where the discussion got interesting was around the topics of automation, and predicting, rather than reacting to, IT problems. Unsurprisingly, this got everyone talking about AI in general and AIOps in particular.
For this small group of IT leaders, technology vendors bearing AI gifts are treated with a high degree of scepticism. Too often, it appears that some of the solutions that claim to use AI, offer little more than automated scripts and rules-based algorithms. Vendor marketing teams came in for a significant level of criticism for the way in which they are misrepresenting the AI credentials of their solutions.
The other significant concern was the lack of transparency about how AI tools arrive at the conclusions they reach or suggestions they make. One anecdote related to a very large IT vendor marketer who, when asked how their AI worked, said “nobody really knows”. Clearly, “black-box” AI solutions represent an unknown risk that will slow down adoption. Everyone agreed that having the ability to see and understand how the AI algorithms work…and being able to provide input to them…would go a long way to speeding up AI adoption.
Automation was seen as, pretty much, a given. But clearly a level a of concern remains where automated remediations are being fed by AI or Machine Learning decisions. There was much nodding of heads when I suggested that trust in automation for remedying IT infrastructure problems was at a similar stage to vehicle automation. Nobody in the room said that would be comfortable getting into a fully autonomous vehicle for a journey across London, but they would happily use Park Assist to get into scarce and tight city parking spaces.
At Bloor, we have a strong focus on Trust. Much of what we discussed last week indicated a certain lack of Trust either in the veracity of marketing claims, or the lack of visibility into the workings of any particular AI model. Short, sharp Proof of Concept trials were seen as a good way of understanding how quickly and easily AIOps tools could be integrated into an existing environment and begin to deliver tangible business benefits. If AIOps is to become a tool for predicting and preventing IT infrastructure problems rather than reactively solving them, getting the Trust of IT and business leaders will be essential.
Our discussion ended on a lighter and unintentionally (at least to start with) note. The term, “selling snake oil” was used about some marketing communications. It was then suggested that a new term should be used to indicate something good. “OK, describe the opposite of a snake then”. “Having legs” was the first very quick answer. “It should be soft and cuddly” said another. At this point you could see the penny dropping as to where this was heading. “Maybe it should have big black eyes”. “Oh, do you mean like a Panda?” “Yes, a big one!” Cue laughter round the table. Obviously, as an analyst, I couldn’t possibly comment.