Guest: Mav Turner (LinkedIn)
Company: Kentik
Show Name: An Eye on AI
Topic: Agentic AI, Observability

Networks are growing faster than teams can keep up with. AI workloads are surging. And traditional observability stacks no longer give engineers what they need: real understanding. In this in-depth conversation, Kentik’s Chief Product Officer, Mav Turner, explains how the company’s new AI Advisor is changing how teams detect, diagnose, and manage network issues.

Modern networks are under unprecedented pressure. Trillions of data points move through global infrastructure every day, and the rise of AI, microservices, distributed traffic, and dynamic cloud architectures has created an environment where dashboards alone are no longer enough. Observability tools can highlight anomalies, but they rarely explain what those anomalies actually mean — or what engineers should do next.

This is the gap Kentik is aiming to close with its new Agentic AI Advisor. In this conversation, Turner walks through the evolution of network analytics, how reasoning models are reshaping operations, and why understanding context is now the differentiator for modern observability platforms.

A decade of data and context

Kentik has been collecting internet traffic data at scale for more than ten years. The platform pulls information from service providers, Fortune 500 enterprises, and networks that operate at some of the largest global scales. This long history created the foundation for Kentik’s AI approach — not by training foundation models on customer data, but by exposing rich, contextual, real-time insight directly to reasoning models.

Turner explains that early uses of AI inside Kentik were closer to chatbot interfaces. A user asked a question, the system generated a dashboard or a visualization, and that was useful, but still limited. What changed recently is the ability to supply these newer reasoning models with deep, contextual knowledge of network behavior. “When you ask a question now, it doesn’t just return a number,” he says. “It explains why something is happening, how it reached that conclusion, and what factors matter.”

This shift from summarizing data to understanding data is the heart of Kentik’s Agentic AI Advisor.

Going beyond chatbot wrappers

The industry has seen a wave of “chat with your logs” features added to observability products. Most are wrappers around an LLM that interpret text, but lack intrinsic understanding of network structures, routing logic, or device relationships. Turner is blunt about it: that approach can be helpful, but it’s shallow.

Kentik’s model is tuned with a detailed semantic understanding of the network itself. It knows which data sources exist, what they represent, and what they can — and cannot — answer. This reduces hallucination risk, a critical requirement when dealing with infrastructure operations. Turner shares an example where the system asked a customer for configuration files because it didn’t have access to them at the time. Instead of guessing, it admitted the gap and explained what it could do with the missing data. That willingness to answer within precise boundaries is what keeps AI safe for network teams.

Why trillions of data points matter

Kentik ingests massive amounts of real-time data every day. But the important part isn’t the raw volume — it’s the context surrounding it. With so many variations in traffic patterns, routing paths, infrastructure types, and edge conditions, the AI Advisor sees patterns that would otherwise remain invisible. It’s not trained on this customer data, but it can use it in real time to reason through situations with more accuracy.

Turner notes that this exposure to diverse network conditions gives the system a long-tail understanding of problems. That means more accurate answers, earlier detection of issues with similarities to past patterns, and more reliable troubleshooting for teams that may not have deep expertise in every part of the network.

Addressing the skills gap

One of the clearest themes in the conversation is the growing disconnect between network scale and available engineering expertise. “I haven’t talked to a team that says they have plenty of capacity,” Turner says. Every organization is being asked to do more with less, and networking is one of the hardest areas to staff.

Kentik’s AI Advisor is designed for augmentation — not replacement. The goal is to take common troubleshooting tasks, ticket resolutions, and first-level investigations and make them accessible to junior technicians or frontline support teams. Instead of escalating issues to senior engineers, frontline staff can use natural language queries inside Kentik to validate whether a problem is related to the network, where it is occurring, and what actions might resolve it. This removes countless interrupts from senior engineers’ workloads and lets teams work at the speed modern networks demand.

According to Turner, the same system also supports advanced engineers, but at a different level. For architects and senior network professionals, the AI can provide insights into long-term design decisions, cost-optimization paths, or better routing strategies. It serves both operational and strategic functions, depending on who is asking the question.

Validated recommendations and realistic autonomy

Trust is a recurring point in the interview. AI can interpret data, but how much autonomy should it be allowed to take?

For Kentik, direct configuration changes are still a human-in-the-loop domain. The company has no interest in letting an AI autonomously reconfigure networks across vendors and environments. The real autonomy, Turner says, is in investigations — proactive detection of conditions that may warrant attention.

This is where agentic behavior emerges. Instead of waiting for a user to ask a question, the system can analyze what’s happening, surface a potential problem, and explain why it matters. A future version might even say things like, “Your peering relationship here is inefficient; shifting traffic could reduce spend.” That kind of proactive advisory capability is where Kentik sees the next evolution of network observability.

Turner emphasizes that the validation step remains critical. Engineers should review why the AI is making the recommendation, understand the data behind it, and make the final call. Transparency is built directly into the system’s design, ensuring engineers never feel like the AI is a black box.

Where agentic AI in observability goes next

The current explosion of AI agents across enterprises has created fragmentation. Every vendor introduces its own agent. Every platform wants to be the hub. But enterprises don’t want 20 different AI entry points. They want one governed, secure interface that can interact with specialized systems underneath it.

Turner expects a consolidation wave, where AI agents become interoperable and orchestrated at an enterprise level. Standards like the Model Context Protocol (MCP) may play a role, but today many vendor implementations are still API wrappers rather than deeply integrated workflows. The long-term picture involves multi-agent collaboration, domain-specific reasoning engines, and shared context across systems like observability, CMDBs, and ITSM platforms.

This is also why partnerships matter. Kentik already works closely with platforms such as ServiceNow to integrate observability insights into remediation workflows. The focus is not creating yet another silo, but fitting naturally into what enterprises already use.

Is there an AI bubble?

The conversation closes on a broader question: are we in an AI bubble?

Turner differentiates between a bubble and a transition. During the dot-com boom, many companies disappeared, but the internet remained and became foundational. The same will happen with AI. Some experiments will be discarded, many tools will consolidate, and the industry will mature. But the underlying shift — using AI to reason about complex systems — is not going away.

Enterprises may move cautiously, especially in regulated sectors, but the consumer side has already pushed the expectation forward. Just like the BYOD movement reshaped enterprise IT a decade ago, everyday AI usage will influence what companies demand from their infrastructure tools.

Closing

Kentik’s Agentic AI Advisor is part of that shift. It represents a new category of observability: one where data is not just visualized, but understood. And as networks continue to evolve, that kind of contextual reasoning may become a requirement rather than an enhancement.