Five Trends Driving the Next Wave of AI and AIOps – JAXenter

The evolution of AI and AIOps today is being driven by five key trends which, while each important on their own, are much more striking when taken together. As a collective it forces a concentration on the kinetic aspects of AI. Let me outline the five biggest trends affecting AI and AIOps today, and explain why IT organizations should track their development and their implications.

SEE ALSO: Machine learning in finance: From buzzword to mainstream

The first trend is widely recognised: DevOps has made a continuously evolving IT environment a reality. IT has always evolved rapidly, a constant feature of the industry. But now, with DevOps adopted in many organizations, were going to another level. By design, weve introduced continuous evolution into our environments, and constancy has been designed out. AI strategies have always assumed there is a degree of constancy in the environment. This is no longer the case. IT environments are now expected to change continuously at all layers of the infrastructure and application stack, and these changes can frequently transpire in a matter of microseconds. Traditional data sampling timeframes, as well as human and robotic reaction times to actual or potential incidents (at best, a matter of seconds), are no longer effective or even informative.

Economic pressures have been exacerbated by the pandemic, so users now have an increased demand for automation of tasks. Until now, AI has been an exercise in the automation of analysis. Today, in almost any business context, users want AI to take the next step and drive automation to make humans more productive, in IT remediation for example. Traditionally, there has been resistance to this. But if you couple increased complexity with new economic realities, automation takes centre stage as AI evolves to drive increased productivity as a key benefit. If an AI solution cant deliver on automation, it will be perceived as incomplete.

The next trend is a consequence of increased complexity. It is now impossible to disentangle security management issues from operations management issues. There is no way of determining whether or not a given anomaly is a consequence of an unpublicised change, a system error, an unexpected change in user behaviour, or a malicious intervention. The attempt to run security and operations management side-by-side from start to finish is on the verge of breaking down. In the very near future, human and automated IT system data analysts will observe and analyse anomalies and surprising patterns purely with regard to the degree of their unexpectedness. The prime directive will be to pick up whats new and surprising as early as possible. The strategy for dealing with the unexpected will only come as a next step.

Remote work, the fourth trend, has been in place for a long time but has been accelerated by the drastic social distancing measures adopted in response to the pandemic. All teams addressing incidents will now be virtual and ephemeral, working on any given situation via a virtual Network Operations Centre. As people work from anywhere, teams will be assembled cross-functionally depending on the problem at hand, and disbanded once a solution is implemented.

Although the five dimensional approach to AIOps advocated by Moogsoft did not anticipate the Covid pandemic per se, it did anticipate the need for supporting an increasingly dispersed, virtualised, and ephemeral ITOSM (IT Operations and Service Management) workforce. The role of AI in collaboration enablement and virtual learning has already moved front and centre for many enterprises, and that role will only become more prominent as the post-Covid new normal becomes established.

Finally, because of continuous change and the requirement for short-term ROI on technology investments, any attempt to build ontologies, service models, or configuration management databases, that have no predictive power, will be seen as useless projects.

Historically, there has been scepticism surrounding such projects that promise to contain a single record of truth. However, given that changes are continually taking place in the environment and given the business need for immediate returns, any modelling that takes place needs to be an engine for prediction, and if its not that, its relatively pointless.

SEE ALSO: How enterprise companies are changing recruitment with AI

Any one of these trends operating individually would go a long way towards putting automation at the center of both AIOps strategies and AIOps technologies. All five of them working together, particularly under the pressure of the recent pandemic, mean that kinetic AIOps will be a critical element of digital business in the very near future. Crucially, any AIOps solution will be seen to be redundant unless it robustly supports the automation of any remedial or preventative actions recommended by the solution.

I will go out on a limb and predict that, in two years time, any AIOps solution with functionalities that are just confined to the first three dimensions (data selection, pattern discovery, and causal inference) and dont support collaboration and automation, will not be acceptable to effective enterprises.