Artificial Intelligence (AI) – once the basis of Science fiction is now a reality. Though it is often accused of being a mere buzzword, it is complicated to overstate its impact on the human experience, and as AI’s integration becomes more global it is necessary to understand its potential. It is important to zero in on what AI has to offer Information Technology and how integrating an AIOps driven IT department is beneficial to an enterprise.
There are many ways to define Artificial Intelligence, but in layman’s terms, the goal of any AI program is to “mimic cognitive functions that humans associate with the human mind such as learning, problem-solving, and decision-making. The term AI encompasses many branches of computer science such as machine learning and evolutionary computation, both very different. Evolutionary computation refers to an algorithm that solves a specific problem by producing possible solutions, picking the best one, in contrast, machine learning refers to a technique used by programs to identify what tasks are needed to be performed based on data.
One of the first high profile AI systems was IBM’s Watson. When Watson was first introduced to the world through its infamous 2011 Jeopardy victory, it was a media sensation, promising world-changing innovation and progress. Part of what made Watson, so revolutionary is the fact that Watson “understood” natural language used by people conversationally. This accomplishment directly led to the speech-based intelligent systems such as Alexa, Siri, and Okay, Google.
Nearly, nine years after Watson’s debut, artificial intelligence has been integrated in most sectors. Though the lens pop culture and the press views AI is one of either sweeping change and significant progress or the perils of automation and deep fakes, there are many low-profile AI systems that people interact with on a daily basis, from online banking to browsing streaming service’s content suggestions, most aspects of human existence in the developed world are enhanced by Artificial Intelligence. A branch of AI that is experiencing significant growth is the AIOps market
AIOps is the use of artificial intelligence for IT operations. Every enterprise depends on a functioning IT department to ensure its success. In the event of a system failure, it is up to the IT department to resolve the issue as fast as possible to ensure that overall efficiency is not affected. When IT is managed solely by people there are a lot of moving factors involved. The typical sequence of IT solving a problem is as follows:
There are many inefficiencies in this sequence. Perhaps the most evident is an over-reliance on a person’s individual knowledge/experience of how the system works. There is no way a person can know all the previous problems in the system and how they were resolved and time can be wasted in trying to solve a problem through trial and error. Second, there can be a communication mishap in communicating the exact problem experienced by the user and therefore lead to an extended breakdown of the system. An AIOps driven system can streamline this process and potentially allow the department to solve problems before any user interruptions occur.
The AI of AIOps refers to deep learning techniques to analyze past data to spot trends in the system and present data to identify current problems that may plague the system. The longer an AIOps system is deployed, the more powerful it becomes because the backbone of any AI system is data, so the more data it receives the more effective it becomes. This is the exact reciprocal of human capabilities, while humans are superior to computers in almost every way they can not process and understand large amounts of data in a timely manner. An AIOps system can only age well.
Integrating AIOps can turn a fledgling IT department into a powerful one.
Not all AI applications are created equal. A badly designed AI can derail an entire operation. An AIOps system is the sum of all its parts therefore all the details that make up an AI affect its overall quality.
Though the user interface of an application doesn’t involve any artificial intelligence programming it is very important to the efficacy of an AIOps system. A poorly designed and unintuitive UI may not hinder the ability of the AI system to properly monitor and diagnose problems but if the UI does not lend itself to accessibility it essentially becomes a lame-duck program. Designing a UI that allows the user to truly appreciate the “intelligence” of the system and navigate the capabilities of the program should be a top priority of designers, conversely, users should familiarize themselves with the UI to understand the breadth of tools available to them and how to best use them to optimize their IT systems.
An AIOps system that cannot convey its message to humans is ineffective. If an AIOps message cannot deliver a notification pointing to a problem in the infrastructure in a concise and informative manner it can cause more problems than solve them. All too often AIOps systems group different problems under a generic name and when one of those problems arises the human monitoring system receives a generic notification, informing them that there is a problem but not what the problem is or how to solve it. This can lead the department down a rabbit hole of trying and failing to find a problem which leads to time wasted and frustration. Another important facet of the messaging system of an AIOps system is conveying which problems are critical and which aren’t. If all notifications are shown to have equal weight than the messages begin to become cluttered and difficult to navigate. Allowing an IT department to prioritize which problems to solve first is essential to maintain order. Again, it can be seen that though the AI programming may be well done if the user-facing functions are not well-designed it renders the “AI” of AIOps moot.
An AIOps system’s report is analogous to allowing users a “peep under the hood” of the AI capabilities. It is a detailed summary of the data associated with the system that the AI is using to spot trends to diagnose problems. A generated report must be filled with relevant information and digestible if not it doesn’t serve its purpose and is basically the appendix of an operation.
The aforementioned problems all have to do with the elements of an intelligent system that aren’t related to the specific AI programming but the user facing elements that can hinder the system’s overall efficacy. However, what truly distinguishes a “good” AI from a “bad” AI is the way that it processes information. A well-designed AI is able to process data and produce information directly from that. A badly designed AI may produce information that is technically correct but can only do so from tangentially related data. This causes problems down the line because if an AI system is given an input that is correct, it builds off of that trend line which can produce inaccuracies in time.