When Ashan Willy, the CEO of New Relic, took the stage at Transform 2025, he brought a fresh\u2014and much-needed\u2014perspective to the AI conversation. Instead of just touting data or breakthrough algorithms, Willy spotlighted something many overlook: observability. His message was simple but hard-hitting: as we build smarter, more autonomous AI systems\u2014what he calls “agentic AI\u201d\u2014we need to make sure we can actually see how they work.<\/p>\n
So what exactly is agentic AI? Think of these as systems that don\u2019t just process data or follow basic instructions; they make decisions by themselves in real time, working toward goals we set for them. But with this kind of autonomy comes complexity. It\u2019s a bit like hiring a super-smart employee who never sleeps and never asks for days off, but who also moves so quickly that it\u2019s difficult to keep track of what they\u2019re doing. That\u2019s why Willy keeps hammering home the need for better visibility\u2014or observability\u2014into their actions.<\/p>\n
Willy puts it plainly: “You can\u2019t improve what you can\u2019t measure.\u201d The more independence we give these AI agents, the more important it is to be able to follow their thought processes, measure their performance, and understand their choices. Observability tools are the answer\u2014they give development teams a real-time window into what these AI agents are doing, how they\u2019re performing, and whether they\u2019re behaving as expected.<\/p>\n
There\u2019s a long-running joke (and an ongoing headache) in tech about the \u201cblack box\u201d problem. We build smart systems, but sometimes even the creators can\u2019t explain exactly how an AI engine arrived at a particular conclusion. This isn\u2019t just a curiosity\u2014it\u2019s a real issue, especially when AI is used in places where mistakes aren\u2019t an option.<\/p>\n
This is where Willy and his team at New Relic come in. Their goal is to change that black box into what he calls a “glass box”\u2014where everything inside is visible, measurable, and manageable from day one. By baking observability into the very core of autonomous AI systems, they\u2019re aiming for a world where decisions can be traced, tweaks can be made quickly, and accountability isn\u2019t an afterthought.<\/p>\n
Beyond transparency and trust, there\u2019s another big reason to care about observability: return on investment. Companies are pouring serious money into AI, and it\u2019s no longer enough to hope things will work out. Observability provides the essential data and context you need to tune performance, streamline operations, and spot trouble before it blows up.<\/p>\n
As Willy puts it, when you can monitor what your AI agents are up to, you can iterate more rapidly, fix problems before they become disasters, and ultimately deliver a lot more value. In other words, observability is what allows teams to move from just reacting to problems to proactively managing them\u2014making AI safer, faster, and more reliable everywhere it\u2019s deployed.<\/p>\n
And as more businesses roll out their own intelligent agents, the need for robust observability will only grow. Companies that make transparency and monitoring a priority now won\u2019t just have better AI\u2014they\u2019ll have a real advantage, building more dependable and scalable systems for whatever comes next.<\/p>\n