IT operations professionals have a lot on their plate, and when an incident occurs that takes down a critical system, it's always a race against time. For years, companies have sought the benefits of getting back on their feet faster with playbooks designed to find answers to common problems and perform post-mortems to prevent them from happening again. But not all problems are easy to solve, there is so much data and so many possibilities. Point of failure.
In fact, this is the problem that generative AI is perfect for solving, and AIOps startup BigPanda today announced a new generative AI tool called Biggy that will help solve some of these problems faster. . What Biggy is designed to do is examine a wide variety of IT-related data to learn how your company operates, compare it to problem scenarios and other similar scenarios, and suggest solutions. is.
BigPanda has been using AI since the company's early days and intentionally designed two separate systems: one for the data layer and one for the AI. This prepared us for the transition to generative AI based on large-scale language models. “AI engines before Gen AI were building a lot of other kinds of AI, but what we're doing with Biggy and feeding into what we're doing with generative AI and conversational AI. ,” BigPanda CEO Assaf Resnick told TechCrunch.
Like most generative AI tools, this tool makes available a prompt box where users can ask questions and interact with the bot. In this case, the underlying model is trained on data within the customer company and publicly available data on specific hardware or software, and is designed to address the types of problems that IT departments regularly deal with. has been adjusted to. .
“The out-of-the-box LLM is trained on vast amounts of data and is actually very good as a generalist for all the operational areas we focus on, including infrastructure, network, and application development. And in fact, they know all the hardware very well,” said Jason Walker, Chief Innovation Officer at BigPanda. “So if you ask about the specific HP blade server where this error code occurred, you can summarize it very well, so many events he uses it for traffic.” Of course, there's more to it than that. must be. If not, a human engineer can easily find it with her Google search.
Combine this knowledge with the ability to filter internally across different data types. “BigPanda ingests customer operational and contextual data from observability, change, CDMB, and topology, along with historical data and human and organizational context, converting data into key-value pairs, So we normalize it to a tag,” Walker said. He said. It's full of jargon, but basically it means looking at system-level information, organizational data, and human interactions to provide responses that help engineers solve problems.
When a user enters a prompt, it examines all the data and generates an answer that can point engineers in the right direction to solve the problem. They acknowledge that generative AI is not always perfect as it does not exist, but they will let the user know when there is low certainty that the answer is correct.
“In areas where we don't think there's a lot of certainty, we tell them this is the best information we have, but it needs to be confirmed by humans,” Resnick says. In other areas where there is more certainty, automation could be deployed and work with tools like Red Hat Ansible to resolve issues without human intervention, he said.
The data ingestion part isn't always easy for customers. This is the first step toward providing an AI assistant that helps IT departments get to the root of problems and resolve them faster. While no AI is foolproof, interactive AI tools should be an improvement over current time-consuming manual approaches to troubleshooting IT systems.