Context is King: Why Network AI Needs Domain Knowledge to Work
Summary
Generic AI fails in network operations because it lacks the “institutional knowledge” of your specific environment and business priorities. Learn how Kentik’s Custom Network Context encodes your unique operational reality into AI Advisor, turning a generic chatbot into a context-aware teammate.
AI has made its way into just about every corner of IT, but network engineers remain understandably skeptical. Maybe it’s because we’re tired of the marketing hype, or maybe it’s because we’ve all seen AI tools that sound impressive in demos yet struggle the moment they encounter a real production network…a network filled with historical baggage, inconsistent naming conventions, institutional knowledge, and business constraints that exist nowhere in the documentation.
That’s exactly why context matters. And it’s why Custom Network Context is one of the most important capabilities of AI Advisor, which we launched just a few weeks ago.
Why generic AI falls short in network operations
Large language models are great at understanding general networking concepts. They’ve been trained on the entire internet, including Cisco textbooks, networking blog posts, Arista forums, Juniper KBs, and so on. They know what BGP is, how packet loss impacts applications, and why asymmetric routing can cause problems. The problem is that they don’t know your network.
They don’t know that “core-dc1-switch-12” is a more critical device than “agg-edge-03.”
They don’t know which IP ranges represent customers versus internal services.
They don’t know that maintenance windows on Sundays mean alerts should be treated differently.
Without that knowledge, AI can still provide answers, but not always useful ones. Kentik’s approach to AI starts with a simple principle: AI should behave like a trained network engineer joining your team, not like a generic chatbot making wild guesses.
What is Custom Network Context?
Custom Network Context is a built-in function of Kentik AI that allows you to provide organization-specific knowledge to the system in plain text or Markdown (up to 100,000 characters) that becomes part of the system context used by AI Advisor across all conversations.
This isn’t a one-off prompt or a temporary chat instruction. It’s a persistent, shared context that applies every time AI Advisor reasons about your network.
You can include things like:
- Network architecture and design details (WAN, campus, data center, cloud, etc.)
- IP addressing schemes and ASN usage
- Device naming conventions and application tagging logic
- Definitions of internal terms (what “customer” or “critical” means to you)
- Maintenance windows and operational constraints
- Traffic priorities and business-critical applications
- Organizational preferences for troubleshooting or escalation
In essence, Custom Network Context teaches Kentik AI how your network actually works.
Where it fits in the Kentik AI ecosystem
Think of Kentik AI not as a single feature but as an entire ecosystem. At the foundation is Kentik’s data platform, which unifies flow, metrics, cloud, synthetic telemetry, and various metadata. On top of that sit AI-powered capabilities like Query Assistant, Cause Analysis, and AI Advisor.
AI Advisor is the agentic layer. It interprets intent, builds investigation plans, selects the right tools, gathers data, analyzes results, and presents conclusions, often with recommended next steps.
Custom Network Context sits alongside Natural Language Runbooks as one way customers tune AI Advisor. Runbooks guide how to investigate known scenarios. Custom Network Context defines who you are and how your network behaves.
Notice the image below that we can simply tell AI Advisor the details that it needs to know, embedding the domain-specific context into the system itself.
Together, Runbooks and Custom Network Context allow AI Advisor to deliver answers that are not only correct but also contextually relevant.
How Custom Network Context improves AI Advisor
The most important impact of Custom Network Context is accuracy through alignment. When AI Advisor investigates an alert or answers a question, it reasons across:
- Live Kentik telemetry
- The tools it has available
- General networking knowledge
- Your custom network context
This dramatically reduces false leads and unnecessary investigations. Instead of suggesting checks that don’t apply to your environment, AI Advisor can focus on what does matter.
It also improves consistency. By including institutional knowledge once, you ensure that every engineer, junior or senior, gets answers aligned with how your best engineers think.
And lastly, it keeps the human in the loop. AI Advisor shows its reasoning and data at every step, allowing engineers to validate conclusions rather than unquestioningly trusting them.
Faster, smarter incident triage for the real world
Imagine an enterprise network in which “Customer traffic” is defined by a specific set of ASNs and BGP communities, certain interfaces are always noisy and should be deprioritized during investigations, Sunday nights are scheduled maintenance windows, and a small set of applications are revenue-critical and must be investigated first.
Without Custom Network Context, AI Advisor can still analyze flows, metrics, and alerts, but it lacks a business lens and the institutional knowledge embedded in the minds of the people on the network team.
Using Custom Network Context, when an engineer asks, “Is this traffic spike impacting any customers?”, AI Advisor knows what “customer” means in your environment. That way, it can filter traffic appropriately, ignore known maintenance noise, focus on the relevant applications, and surface impact in business terms, not just packets and flows.
In the following image, notice we’ve added customer-specific information so that AI Advisor understands what we mean and can create the most appropriate and effective workflows.
What used to take 30 minutes of manual setup across dashboards now happens in seconds, with reasoning displayed for review and all the data attached to the result.
Why this matters for real networks
Networks today are more complex than ever, including hybrid, multi-cloud, overlays, and SaaS, and are constantly changing. At the same time, teams are expected to do more with fewer people and smaller budgets. AI helps only if it reduces cognitive load instead of adding uncertainty.
Custom Network Context ensures that Kentik AI doesn’t just understand networking. It understands your network. It encodes your environment, your priorities, and your operational reality directly into the AI system.
That’s what turns AI Advisor from a clever chatbot weekend project into a trusted teammate. And that’s the difference between AI that looks good in theory and AI that actually works in production.
Want to see Custom Network Context in action? Set up a personalized demo of AI Advisor today.
