Grokstream has updated its Grok artificial intelligence for IT operations (AIOps) platform to better identify patterns and prioritize incident responses.

Company president Josh Kindiger said the Proactive Problem Identification capability makes use of predictive and causal artificial intelligence (AI) models to now identify clusters of related recurring anomalies, determine the root cause and surface automation recommendations to prevent them in the future.

Additionally, the company has added GrokGuru, a generative artificial intelligence (GenAI) tool to capture tribal IT knowledge that can then be used to generate human-readable summaries and recommendations.

Finally, Grokstream has added a no-code interface to GrokConnect to make it simpler to ingest, transform, enrich, and shape data from third-party monitoring, observability, service management, and infrastructure tools and platforms.

The overall goal is to improve application availability and performance by relying more on machines to manage complex IT environments, noted Kindiger.

It’s not clear how broadly IT teams are adopting AIOps platforms, but as IT environments continue to become more complex, they are becoming all but impossible to troubleshoot without some help from advanced analytics. Grokstream is making a case for a platform that combines multiple types of predictive, causal and now generative AI models to both triage incidents and identify best practices to prevent as many of them as possible from occurring in the first place.

Precisely which tasks will be handled by AI agents and models versus a human is still very much a work in progress. However, ultimately most IT tasks will eventually be performed by AI agents that will obviate the need to have a dedicated IT operations team, said Kindiger.

Between now and then, however, IT teams through trial and error will still need to determine how much confidence they have in AI platforms to manage workflows that are, in the main, deterministic in the sense they need to be completed the same way each time. Generative AI models, in contrast, are probabilistic, which means they almost never perform the same task the same way twice. While there are plenty of tasks, such as generating a report, where a probabilistic outcome might suffice, there are also just as many tasks where precisely how the task was executed matters greatly.

In addition, there are times when an AI agent for some unexplained reason will report it has completed a task when in fact it has not.

Of course, the worst AI any organization will ever have is the one they have today. The pace of AI innovation, especially when it comes to reasoning, only continues to accelerate. In fact, many IT teams are already making extensive use of AI agents to create scripts and generate reports. However, taking AIOps to the next level will require an integrated platform that makes sure the right type of AI model is being applied to the right task.

In the meantime, IT teams should be creating an inventory of the tasks they can confidently assign to an AI agent today with an eye toward continuing to expand it in the months ahead. In more instances than many of them might not initially appreciate, much of that list is going to be made up of tasks that no one in IT especially enjoys doing in the first place.