Transform 2025: Why observability is critical for AI agent ecosystems

Independent software revolution is coming. At Transform 2025, Ashhan Willie, CEO of New Relic and Sam Witteveen, CEO and co -founder of Red Dragon AI, talked about how they have created the return on measurable investment and draw a road infrastructure map to increase AI Agenic.

New Relat provides the ability to observe customers by picking up and linking applications from a real -time brown. The monitoring capacity goes beyond – it relates to preparing the difference with the context and insight needed to understand, explore and improve complex systems, even in the face of unexpected issues. This day is now much more complicated in the mixture. The company’s observation now includes monitoring everything from NVIDIA NIM, Deepseek, Chatgpt, etc. – AI is almost 30 % use of artificial intelligence monitoring by almost 30 %, a quarter of a quarter, reflecting the acceleration of adoption.

Willie said: “The other thing that we see is a great diversity in models.” “Companies started with GPT, but they started using a full set of models. We have seen about 92 % in the variation of the models that are used. We started seeing institutions to adopt more models. The question is, how can you measure the event?”

Observation in the world of an agent

In other words, how does the observation of the observation develop? This is a big question. Using situations vary greatly through industries, and jobs vary mainly for each individual company, depending on the size and goals. A financial company may focus on maximizing profit margins before benefits, taxes, depreciation and consumption, while a company that focuses on speed products in addition to quality control.

When New Relat was founded in 2008, the center of gravity for observation was to monitor the application for SAAS, the mobile phone, and then the cloud infrastructure at the end. The ascension of artificial intelligence and AICEC AI re -submit the observation to the applications, where agents, micro -sticks and nanoparticles operate the acquired code from artificial intelligence.

Artificial intelligence of the note

With the high number of precise services and services, especially for the original institutions, is digital, the cognitive load of any overwhelming human processing tasks becomes. Of course, Amnesty International can help this, Willie says.

“The way you will work is that you have enough information as you will work in the cooperative situation,” I explained. “The promise of the agents in the observation is to take some automatic work burdens and make them happen. This will lead to his addiction to more people.”

Observing the agent is one platform

One note platform benefits from the agent. The agents automate the workflow, but they constitute a profound integration in the entire ecosystem, through all the multiple tools that the institution has to play, such as harnessing, GitHub, Servicenow, etc. With Alectic AI, developers can be alerted to what is happening with code errors anywhere in the ecosystem and fix them immediately, without leaving their coding platform.

In other words, if there is a problem with the code scattered in GitHub, then the observation platform is supported by the agents discovered, determining how to solve it, then alert the engineer – or completely automated the process.

Willie said: “Our agent is mainly looking for every part of the information we have on our platform.” “This can be anything of the application’s performance, and how to perform the AZURE or AWS basic structure – anything we think is related to spreading the code. We call it the skills of the agent. We do not rely on a third party to know the applications programming facades, etc.”.

In GitHub, for example, they allow the developer to know when the code is well, where errors are dealt with, or even when the software decline is necessary, then they automate this decline, with the approval of the developer. The next step, announced by New Relat last month, works with Agent Copilot Coding to tell the developer exactly about the lines of the code that sees the problem with him. Then Copilot returns, corrects the problem, then gets a ready -made version again.

The future of the agent artificial intelligence

Willie says that organizations adopt the agent artificial intelligence and begin to adapt to them, so you will find that the ability to observe is a decisive part of its functions.

He says: “When you start building all these integrations and pieces of agents, you will want to know what the agent does.” “This is a kind of thinking for infrastructure. Thinking to know what is going on in your production. This is what the observation will bring, and we are at the forefront.”