Analytics and AI/ML were hot topics at our OpenText Operations Bridge Summer Series, so I followed up with Lars Rossen, our chief architect, and Adam Luciano, director of AI and incubation, to delve deeper. Here’s what I asked and what I learned.

What role can AI play in managing enterprise complexity?

As organizations grow, so do their infrastructures. And growing infrastructures sprout more tools and applications that must be monitored and supported. The more advanced AI gets, the more it can help organizations manage that complexity.

According to Lars, the interconnectedness of systems adds to the complexity. He said that back in the day, it was easy to figure out why an application had stopped working. Suddenly the server just wouldn’t answer the application. Today, architectures are so complex that experts are needed. But experts are expensive—you can only hire so many of them. That’s where AI comes in.

One function that uses AI to simplify complexity is Automatic Event Correlation (AEC). AI-based AEC doesn’t require you to manually write correlation rules. It automatically learns the patterns and groups events, reducing alert noise as it guides you to the problem.

So even though technology caused today’s complexity, I have hope that it can also save us from the complexity it created.

Should we trust large language models?

Not surprisingly, there’s a vague mistrust in large language models (LLMs). Lars pointed out that they’re good at making conversation based on lots of data, but they don’t necessarily speak intelligently. So you need to be careful about applying this kind of AI technology. If you are, there’s great opportunity to be had.

What’s an example of how LLMs are used in IT Operations?

We all want a patient support agent that knows how to fix problems quickly. But it’s expensive to have highly trained people at the Level 1 service desk. Lars said that LLMs can support the Level 1 service desk by quickly giving users the information they need in a conversational manner.

A recent study of the impact of LLMs found that contact center agents with access to an AI assistant were 14% more productive, with low-skilled workers improving the most. Customer satisfaction improved too, suggesting that the AI model helped workers better match a customer’s problem to the right business unit for a solution.

What is multivariate detection and why is it needed?

Adam Luciano said we see anomalies in IT Operations every day because we’re managing systems that usually contain thousands of nodes. With small anomalies happening all the time, you could be getting thousands of anomaly alerts a day. It’s hard to know which anomaly is worth looking into. We can harness the power of AI to identify the historical relationships that exist within those different systems.

For example, let’s say there’s an anomaly that’s impacting the response to an API call. Depending on what that API is talking to, that could cause a big service degradation. Multivariate anomaly detection gives you greater confidence that the service degradation is associated with an anomaly.

How does multivariate anomaly detection work?

Multivariate anomaly detection functionality clusters information from different systems and stores it within the OpenText OPTIC Data Lake. Multivariate anomaly detection algorithms then determine existing patterns and relationships from different systems and bring them into our event workflow process. Over time, you gain a contextual explanation of the anomaly.

Ultimately, multivariate anomaly detection cuts out the alert noise and enables you to effectively prioritize anomalies when they appear.

This post is just a small taste of the topics we covered in the Operations Bridge Summer Series. From now until August 31, you can enjoy a full-course meal of on-demand content right here.