Hybrid networks have outgrown the dashboards built to monitor them. The industry’s response of bolting generative AI onto existing tools has produced a wave of “copilots” that, too often, add a chat box to the same alert noise. Selector is making a different bet: that AI-native observability, built from the ground up around correlation, language, and action, is the platform layer on which the next decade of operations will run. At AI Field Day 8, held in San Jose, California, this past May, I was offered an up-close and personal look at how Selector is placing its bet, who it is pressing on, and where the technology is likely to go next. Here’s my read.
The Network Observability Problem Has Outgrown the Network Engineer
For most of the past twenty years, network observability has been a NetOps problem. Engineers stared at colored dots on dashboards, traced packet captures, and triaged incidents in the order they were noticed. That model has quietly broken. Hybrid networks now stitch together on-premises infrastructure, multiple public clouds, regional interconnects, SaaS-delivered services, and ephemeral container fabric, and the topology rearranges itself faster than humans can document it.
What has changed alongside it is the buyer. Application managers, SREs, and, increasingly, CIOs are held accountable when uptime slips, but they typically lack the network-language fluency to interpret legacy telemetry. The industry has been racing to close that gap, first with broader observability platforms, then with AIOps overlays, and most recently with LLM-powered “copilots” layered on top of existing stacks. None of those approaches has fully landed because they all preserve the same underlying assumption: that humans would do the hard correlation work. The more interesting bet is that the platform itself should do it.
That bet also sits within a market facing real headwinds. CSOs across regulated industries have grown wary of exposing operational telemetry to public LLMs. Recent academic work has documented prompt-injection and data-exfiltration attacks against production AI systems, and engineering leaders openly say trust in AI-driven automation is low. Any AI-native observability vendor entering this market has to win on both intelligence and trust. Historically, those two qualities have been in tension.
Selector and the AI-Native Bet
Selector was built on the idea that AIOps shouldn’t be a feature attached to a monitor — it should be the monitor. The company’s stack ingests logs, metrics, configs, flows, and topology from more than 300 sources, then runs them through a patented multi-domain correlation engine that surfaces root causes across the network, infrastructure, and application layers without rule tuning. Above that engine sits a domain-specific Network Language Model and an operational digital twin that let teams ask questions in plain English and simulate change before it happens.
The combination matters because it inverts the standard observability flow. Instead of an engineer pivoting across dashboards to assemble a narrative, the platform delivers the narrative and surfaces the next safe action. That same model lets the company punch above its weight with Fortune-tier customers (including three of the Fortune 5 across financial services, healthcare, and retail, alongside global telcos like Bell, Singtel, Lumen, and TracFone) without relying on the multi-year integration cycles a legacy NMS would require.
Where Selector Lands in the Market, and What It Pressures
In practical terms, Selector’s platform fits where the industry’s pain is sharpest right now: hybrid environments where the network is the most volatile variable and traditional tools throw alerts faster than humans can read them. The product specifics (the AI correlation engine, the Network Language Model copilot, the operational digital twin, and an action layer that integrates with ServiceNow, Slack, Teams, and CLI) align with the workflow of the actual on-call engineer, not the workflow assumed by product marketing. And because the underlying architecture is LLM-agnostic, the same platform can run against a customer’s internally hosted foundation model when CSO requirements demand it. This is a structural answer to one of the biggest blockers to enterprise AI adoption in the field today.
What that posture means for adjacent fields is more interesting than the head-to-head. Selector’s correlation reach extends into application performance, SecOps, and SRE practice; the action layer edges into incident-management vendors’ territory; and the digital twin overlaps with the change-modeling tooling adjacent to network automation. Competitors who built their AI roadmaps as features atop legacy observability (e.g., Datadog’s Bits AI agents, Dynatrace’s Davis AI, and Cisco’s Splunk-and-ThousandEyes consolidation) now face a competitor whose architecture starts with AI rather than ends with it. The most likely response from platform incumbents is acquisition or a deep partnership, not a greenfield rebuild. Looking further out, the category’s trajectory points toward AI-native consolidation over the next three to five years, with agent-in-the-loop validation and uncertainty disclosure becoming table stakes for enterprise procurement.
Conclusion
Selector is one of the more credible answers to a problem most networking vendors still treat as a UI exercise. The product is built where the buyer actually lives: in the chaos of hybrid environments, with real telemetry, real incidents, and real engineering trust to earn. My read: it is well positioned for now, and the technology choice (AI as a platform, not a feature) is the kind of bet that ages well as the rest of the market catches up.
To see how it works in your environment, the Selector team invites you to request a demo at selector.ai.
