The autonomous software program revolution is coming. At Rework 2025, Ashan Willy, CEO of New Relic and Sam Witteveen, CEO and co-founder of Crimson Dragon AI, talked about how they’re instrumenting agentic techniques for measurable ROI and charting the infrastructure roadmap to maximise agentic AI.
New Relic gives observability to clients by capturing and correlating utility, log, and infrastructure telemetry in actual time. Observability goes past monitoring — it’s about equipping groups with the context and perception wanted to grasp, troubleshoot, and optimize complicated techniques, even within the face of surprising points. Immediately that’s turn out to be a significantly extra complicated enterprise now that generative and agentic AI are within the combine. And observability for the corporate now contains monitoring all the pieces from Nvidia NIM, DeepSeek, ChatGPT and so forth — use of its AI monitoring is up roughly 30%, quarter over quarter, reflecting the acceleration of adoption.
“The opposite factor we see is a large range in fashions,” Willy mentioned. “Enterprises began with GPT, however are beginning to use a complete bunch of fashions. We’ve seen a couple of 92% enhance in variance of fashions which can be getting used. And we’re beginning to see enterprises undertake extra fashions. The query is, how do you measure the effectiveness?”
Observability in an agentic world
In different phrases, how is observability evolving? That’s a giant query. The use instances range wildly throughout industries, and the performance is basically totally different for every particular person firm, relying on measurement and objectives. A monetary agency is perhaps centered on maximizing EBITDA margins, whereas a product-focused firm is measuring pace to market alongside high quality management.
When New Relic was based in 2008, the middle of gravity for observability was utility monitoring for SaaS, cellular, after which ultimately cloud infrastructure. The rise of AI and agentic AI is bringing observability again to functions, as brokers, micro-agents, and nano-agents are operating and producing AI-written code.
AI for observability
Because the variety of companies and microservices rises, particularly for digitally native organizations, the cognitive load for any human dealing with observability duties turns into overwhelming. In fact, AI might help that, Willy says.
“The way in which it’s going to work is you’re going to have sufficient info the place you’ll work in cooperative mode,” he defined. “The promise of brokers in observability is to take a few of these automated workloads and make them occur. That can democratize it to extra folks.”
Single platform agentic observability
A single platform for observability takes benefit of the agentic world. Brokers automate workflows, however they type deep integrations into your entire ecosystem, throughout all of the a number of instruments a company has in play, like Harness, GitHub, ServiceNow, and so forth. With agentic AI, builders might be alerted to what’s taking place with code errors wherever within the ecosystem and repair them instantly, with out leaving their coding platform.
In different phrases, if there’s a problem with code deployed in GitHub, an observability platform powered by brokers can detect it, decide how one can clear up it, after which alert the engineer — or automate the method solely.
“Our agent is basically taking a look at every bit of knowledge we’ve on our platform,” Willy mentioned. “That might be something from how the appliance’s performing, how the underlying Azure or AWS construction is performing — something we expect is related to that code deployment. We name it agentic expertise. We don’t depend on a 3rd occasion to know APIs and so forth.”
In GitHub for instance, they let a developer know when code is operating advantageous, the place errors are being dealt with, and even when a software program rollback is important, after which automate that rollback, with developer approval. The following step, which New Relic introduced final month, is working with Copilot coding agent to inform the developer precisely which traces of code it’s seeing the difficulty with. Copilot then goes again, corrects the difficulty, after which will get a model able to deploy once more.
The way forward for agentic AI
As organizations undertake agentic AI and begin to adapt to it, they’re going to search out that observability is a vital a part of its performance, Willy says.
“As you begin to construct all these agentic integrations and items, you’re going to need to know what the agent does,” he says. “That is kind of reasoning for the infrastructure. Reasoning to search out out what’s happening in your manufacturing. That’s what observability will convey, and we’re on the forefront of that.”
