It can be convenient to think of software as a black box - inputs go on one side, outputs on the other - but what happens when something goes wrong? Where did it fail, why did it fail, and what else happened? Observability is the name given to special tools that help you see inside the black box to figure it all out.

However, not all observability tools are suited for all purposes. With DevOps, DevSecOps, AIOps, all kind of ops, you really need something to cover it all - "x"Ops, say. Well, according to LogicMonitor GM AI Karthik Sj, LogicMonitor is that very tool. It's an xOps hybrid observability platform powered by AI.

Sj is still in his first six months at LogicMonitor, but is no stranger to the inudstry or to AI. He spent 15 years at SAP and most recently came from a GenAI startup in Palo Alto. He saw an exciting opportunity at LogicMonitor, and made time in his busy schedule to tell iTWire about it.

"AI has been here for a long time," he said, "but it didn't catch popular imagination until ChatGPT. It democratised the tech with millions of people."

"Enterprises have been using statistical AI, the evolution of rule-baed heuristics, over the last 10 years. There's been natural language processing for some time. Nobody was excited to talk to those. They'll do some of the job but most people say 'let me talk to a real person'," Sj said.

When it came to generative AI, "it's the 'T' in GPT - the transformer architecture - that really exploded innovation. It's opened a gigantic model and ChatGPT has given an interface to play with that model. It's really opened up a whole rennaissance of companies trying to take advantage of it."

"LogicMonitor has been doing AI for a long time, but our new EdwinAI has infused GenAI into the platform. We've also added new workflows," Sj said.

"There's a complexitiy in tools out there. It's complex and what's best-of-breed depends on which part of the stack you are looking from," he said. "There's application monitoring, infrastructure monitoring, and all kinds of issues happening that need a common context."

"If I have to troubleshoot, don't make log into 20 different tools," Sj said. "It's where engineers get hamstrung. They need a single unified pane, and this is where LogicMonitor EdwinAI and GenAI come in."

"EdwinAI can take an alert and decipher it. It might be a super-technical alert that only an L3 network engineer could normally read. Now GenAI can determine maybe it's an SD-WAN issue at a specific location. Everyone can read it, because GenAI helps package it up as a human readable summary."

"The feature helps get a lot of teams on the same page without needing an expert degree in network operations," he said. Ultimately, what LogicMonitor has achieved valuably lowers the barrier of entry to faultfinding through observability. It can make arcane messages human readable, suggest root causes, and draw together information from numerous places. "It saves a bunch of time," Sj said.

"With hundreds of thousands of alerts, there's high value, solving high pain points."

The EdwinAI that LogicMonitor has introduced was the inspiration of creative people within the business. "They looked at how to define modern-day observability. It was named after Edwin Hubble, because the Hubble telescope redefined modern cosmology."

LogicMonitor's goal with EdwinAI is to arm every enterprise with modern observability, and act like your "10x teammate," Sj said. "AI won't replace people, but it will sit side-by-side."

One example of a customer using EdwinAI in production is Syngenta. "The value was instant," Sj said. "They saw a bunch of alerts come in. Which ones are noise vs. signal? EdwinAI correlated the signals and out of thousands of messages said, 'look at these, here's the context'. They saw a 95% reduction in alert noise."

Syngenta also reported a 30% reduction in IT Service Management (ITSM) incidents, 60% reduction in Mean time to repair (MTTR), and 20% improvement in operational efficiency.

"Within an hour in production it did its work," Sj said. With other products taking six months to get operational, EdwinAI offers a huge uptick in return on investment.

Two years since ChatGPT hit the web, a lot of innovation has happened. "We're in the early stages of GenAI 2.0," Sj said. "Version 1.0 was LLMs, but 2.0 is all about AI agents. You don't want to rely on specific LLMs, instead agents take it even further. It not only answers messages, but can take action. Not only can the agent tell you the problem on your router, it can fix it."

"Our Edwin architecture is on the forefront of this."

LogicMonitor has been experimenting on its own models, and its vast stores of metrics, traces, and event data. "We're ahead of companies starting two years back but only relying on GPT output and GPT wrappers," Sj said. "They're nice for Shakespeare, but not for analysing logs. We wanted something more purpose-built."

It's here that LogicMonitor had its appeal to Sj. "What attracted me to LogicMonitor is solving these pain points and its relevance," he said. "This is hard stuff, making it a good use case. People don't want to read alerts. If you can solve this problem you get high stickiness, and a high reward. It's something people want to pay for."

"It was important for me to find strong return on investment," Sj said. "LogicMonitor has high confidence and trust. People expect LogicMonitor to solve issues. I wanted a company doing this and I'm excited by it."

In addition, Sj says the future will hold GenAI native apps - "like search was a new way and people embraced it, I believe GenAI will create new workflows."

"Think about documents, logs, or calls that people sit on for hours. Someone asks, 'do you remember this issue from two years ago?'," Sj asks. "The answer is no, but imagine if you could harvest all the insight from boardroom calls. This is the opportunity here. There's hidden data, and unstructured data locked away that you couldn't touch and now suddenly it's available. It widens the view of the data you can see."

"That's what I'm excited about."