AI Innovation

There is little doubt at this point that artificial intelligence (AI) is going have a profound impact on every aspect of IT service management (ITSM) and service operations management, but getting the most value out of AI requires, ironically, a deeper appreciation for the way humans approach problem-solving. It’s not enough to throw AI models at IT problems. Instead, there needs to be a composite framework that makes it easy to understand how to apply various types of AI models in a way that everyone involved can understand and, just as importantly, have enough confidence to trust.

A composite Al framework integrates multiple Al models to create a more comprehensive and robust set of capabilities that complement each other. The advantage of composite Al is that it leverages the strengths of different types of AI models optimized for different domains to surface more accurate, actionable insights. Much like the way the human brain works, different types of AI models are essentially lobes that focus on specific tasks to provide IT organizations with reliable recommendations, suggestions and summaries of events. In a similar fashion, AI models process signals from telemetric data that tell them what’s going on, how the users are pressuring the system and how the infrastructure is responding in a way that can be clearly understood by an IT team.

In fact, without that visibility, many organizations may instinctively hesitate to embrace AI simply because of a lack of trust. That’s unfortunate, because it also means that the current level of complexity that makes managing IT at scale so challenging today is not going away as quickly as it should.

At BMC, we’re committed to not just adding AI models to our portfolio but also providing a level of transparency that ultimately demystifies AI and increases trust in the technology.

A Brief AI History

While generative artificial intelligence (AI) will continue to claim its rightful place in the spotlight in the coming year, it will ultimately be the mixing and matching of multiple types of AI models that will determine how quickly and how completely various processes will be automated.

AI models come in three distinct types. Machine learning algorithms have been used for years to create both predictive and causal AI models. Predictive AI models use machine learning algorithms to determine the probability of specific outcomes based on data collected from past events. They are already widely used in various IT operations platforms to determine, for example, the likelihood a service might run out of capacity as the amount of data being processed continues to increase.

Though it is just as powerful, the second type of AI model, causal AI, is employed with less frequency, and uses machine learning algorithms to determine the cause of a complex series of events. Casual AI’s goal is to provide insights into the root cause of, for example, an IT incident to enable IT teams to prevent it from happening again.

Finally, there is generative AI, which uses machine learning algorithms to create a large language model (LLM) that generates content in response to prompts from an end user. The goal is to automatically create text, images and code based on the data used to train the LLM that can then be prompted to generate content. The best-known instance of generative AI is the ChatGPT service created by OpenAI, but there will soon be thousands of LLMs that will be trained using narrower sets of data to create content for specific domains. The expectation is that LLMs trained using narrower sets of data will, in addition to being easier and less costly to build, be less likely to hallucinate because the data sets used to train them are smaller and more thoroughly vetted.

All types of AI models are probabilistic in the sense that they rely on what has preceded to predict the next logical outcome in a sequence. None of these types of AI models supersedes the need for the others. In fact, organizations will soon find themselves using these models in concert with one another to optimize a wide range of processes. For example, organizations can use generative AI to enhance service operation management applications using natural language conversations to make recommendations for case resolution.

At their core, BMC platforms use graph technology to federate multiple AI models in a way that makes the analytics surfaced by various types of AI models easier to comprehend. Instead of merely identifying the root cause of an issue, the BMC approach adds the context needed to prioritize incident response based on the impact they will actually have on the business as time-series metrics are continuously collected.

The BMC Helix composite Al framework achieves that goal by orchestrating AI models that enable sensory reasoning and knowledge-based action planning. Rather than having to construct AI models themselves, the BMC approach provides all the benefits of AI within a framework that can be updated as additional AI advances are made.

Transparency Equals Trust

As Arthur C. Clarke famously observed, “Any sufficiently advanced technology is indistinguishable from magic.” While AI is certainly advanced, it is, at its core, an exercise in using models and algorithms to determine probabilities. In the case of IT, the goal is to collect metrics related to any unexpected event to enable us to determine the probability that a specific series of automated tasks will be remediated.

Each IT organization will need to decide for itself how much to rely on that automation based on the potential risks to the business, but in time, most routine tasks will be handled by machines in much the same way humans reason today. The challenge and the opportunity now is determining how much trust to place in AI solutions when considering the context; for example, resolving an IT support request versus deploying a complex cloud-native application.

One way or another, however, the toil associated with managing IT today is about to be greatly reduced, thanks to AI. The application of AI technology has implications for everything from how IT teams are staffed to the pace at which digital services that advance strategic business goals can be rolled out. The sooner everyone involved understands how AIOps really works, the better off everyone involved with IT on any level will become.