Transform 2025: Why observability is critical for AI agent ecosystems

The autonomous software revolution is coming. Bee Transform 2025Ashan Willy, CEO of New Relic and Sam Witteveen, CEO and co-founder of Red Dragon AI, spoke about how they map out instruments for measurable ROI and mapping the infrastructure road route map to maximize agent AI.

New Relic offers clients perceptibility by capturing and correlating the application, log and infrastructure eletry in real time. Observability goes beyond monitoring – it is about equipping teams with the context and insight that is needed to understand, resolve and optimize complex systems, even in the light of unexpected problems. Nowadays that has become a considerably more complex company now that generative and agent AI are in the mix. And perceptibility for the company now includes monitoring everything from Nvidia Nim, Deepseek, Chatgpt, etc. – The use of the AI ​​monitoring is approximately 30%, quarter than a quarter, as a result of the acceleration of adoption.

“The other we see is a huge diversity in models,” said Willy. “Companies have started with GPT, but are starting to use a whole series of models. We have seen an increase of 92% in variance of models that are used. And we are starting to adopt more models. The question is, how do you measure the effectiveness?”

Observability in an agent world

In other words, how does the perceptibility evolve? That is a big question. The use cases vary enormously in the industry, and the functionality is fundamentally different for each individual company, depending on the size and goals. A financial enterprise can be aimed at maximizing the EBITDA margins, while a product-oriented company measures speed on the market in addition to quality control.

When Nieuw Relikwie was founded in 2008, the center of gravity was for perceptibility application monitoring for Saas, mobile and then ultimately cloud infrastructure. The rise of AI and Agentic AI brings perceptibility back to applications, because agents, micro-agents and nano agents are active and produce AI-written code.

AI for perceptibility

As the number of services and microservices increases, especially for digital native organizations, the cognitive burden becomes overwhelming for every human handling perceptibility tasks. Of course AI can help, says Willy.

“The way it will work is that you have enough information where you will work in cooperative mode,” he explained. “The promise of agents in perceptibility is to take some of those automatic workloads and let them happen. That will democratize it to more people.”

Some platform agent persistability

A single platform for perceptibility uses the agent world. Agents automate workflows, but they form deep integrations throughout the entire ecosystem, in all the multiple tools that an organization has in the game, such as Harnas, Github, ServiceNow, and so on. With Agentic AI, developers can be warned of what is happening with code errors everywhere in the ecosystem and repair them immediately, without leaving their coding platform.

In other words, if there is a problem with code that is implemented in GitHub, a detecting observation platform can be detected, determining how to resolve and then warn the engineer – or fully automate the process.

“Our agent fundamentally looks at every piece of information that we have on our platform,” said Willy. “That could be all of how the application performs, how the underlying Azure or AWS structure performs – everything we think is relevant to that code implementation. We call it agent skills. We don’t trust a third party to know APIs and so on.”

In Github, for example, they let a developer know when the code is performed well, where errors are handled, or even when a softwareback is needed, and then automate it, with approval from the developer. The next step, which New Relic announced last month, works together with Copilot Coding Agent to tell the developer exactly with which lines code he sees the problem. Copilot then goes back, corrects the problem and then prepares a version to re -implement.

The future of agent AI

While organizations accept agentic AI and start adapting to it, they will discover that truthfulness is a crucial part of his functionality, says Willy.

“While you are starting to build all these agent integrations and pieces, you want to know what the agent is doing,” he says. “This is a kind of reasoning for the infrastructure. Reasoning to find out what is going on in your production. That is some observability, and there we are paramount.”