Ask EDA: AIOps built for network engineers

What if your network could talk? What if troubleshooting didn’t mean wading through CLI outputs and dashboards? What if network alarms could explain themselves?

Nokia Event-Driven Automation (EDA) is turning these questions into real features. We just announced an AIOps-driven future for data center networking, and the future is bright for network teams. 

Why network teams need AIOps

Network teams face growing complexity, shrinking error margins and escalating demands from cloud, AI and edge workloads. Traditional monitoring tools and manual troubleshooting are leaving them buried under logs, alerts and dashboards that describe symptoms, not causes.

Enter AIOps.

By continuously analyzing telemetry, understanding intent and surfacing what matters, AIOps enables network teams to move faster, avoid downtime and make confident decisions. It shifts NetOps from reactive firefighting to intelligent, proactive control.

EDA brings AIOps to life by giving network teams the insights, guardrails and tools they need to make better decisions, faster.

AIOps that speaks network: Meet “Ask EDA”

EDA’s new capabilities center on Ask EDA, a natural language interface powered by agentic AI. It's your always-on, always-expert digital network engineer. 

When a switch becomes unreachable, you don’t frantically search logs, metrics and alarms. You just Ask EDA: “I can’t reach spine1. Can you investigate?”

In seconds, EDA checks reachability, correlates logs, parses telemetry and provides an answer: “A BGP neighbor configuration was deleted. This caused dropped sessions and rejected connections.”

That‘s observability and explainability, the real value AIOps has always promised. 

From reactive to proactive: The power of agentic AI

Unlike static, polling-based automation scripts, EDA’s AIOps engine uses a declarative, event-driven model.

The system surfaces anomalies and explains what broke, why it happened and how to fix it. It’s all enabled by EDA’s agentic architecture, which dispatches autonomous digital “teammates” to gather context, correlate events and return with precise, actionable insights.

This is AI fused into the operational core. When you’re troubleshooting with Ask EDA, it might respond this way: “Would you like me to help you investigate this issue?” It engages with operators, parses telemetry, correlates alarms, offers insights and provides solutions faster than a human can. 

Create dashboards through conversation

EDA also introduces a no-code dashboard designer. Instead of navigating menus, you Ask EDA: “Create a dashboard with all the active alarms.” Seconds later, a fully rendered visualization appears.

EDA’s dashboard builder leverages a composable system of telemetry queries and visualization primitives. The AIOps engine transforms your natural language query into a structured dashboard, which automatically renders the data in a visual context.

From visibility to insight to action

EDA’s AIOps understands the why behind changes and alarms. In the past, tools just told you something was wrong. It was up to you to parse the torrent of data available and try to determine what was happening. Ask EDA quickly identifies the problem, describes the root cause and provides a solution you can test in EDA’s digital twin before pushing it to production. 

Here’s a real-world example: A BGP neighbor is removed as part of a scheduled change. Ask EDA tracks the config delta, the control-plane effects and where routing sessions dropped and explains the causal chain. All in seconds.

This is the leap from “data collection” to “network storytelling.” Ask EDA tells you what happened, how it happened, what’s likely to happen next and, most importantly, how to fix it fast. 

Built for engineers, not AI researchers

EDA is built for real-world NetOps and supports a broad range of industry-standard integrations, like ITSM, OpenStack, VMware and Git-based CI/CD pipelines. Teams can easily leverage EDA without reengineering their environment.

EDA also supports versioned rollback and “atomic” transaction commits. It applies changes transactionally across the network fabric. Network-wide transactions roll back automatically if the configuration is rejected by any node. Users can run post-checks manually or with EDA and roll back in one click. That’s zero-downtime operations backed by formal guardrails.

EDA supports full integration with CLI-based workflows, too. Engineers can toggle between intent-driven GUIs, natural language prompts and raw CLI. This flexibility means EDA adapts to team maturity, not the other way around.

Trust, but pre-validate: The role of EDA’s integrated digital twin

One of EDA’s critical innovations is a fully integrated, always-on digital twin that continuously mirrors the live network's state and intent and validates every change initiated through EDA.

Before anything touches production, the system runs pre-checks in a virtualized replica of the network fabric. It verifies intent, checks for errors and ensures that the change won't violate policy or introduce instability. If validation fails, the operator is alerted with clear, explainable details. This helps eliminate network outages caused by human error.

This digital simulation architecture is especially powerful for AI-powered operations. Generative AI can make recommendations, interpret natural language queries and suggest changes, but sanity-checking AI outputs with a digital twin brings confidence and safety. 

EDA’s digital twin acts as a guardrail against LLM hallucinations. Any network changes proposed by LLMs can be tested for safety in the digital twin before pushing them to production.

By combining generative AI with deterministic pre-validation, EDA ensures that operator trust in automation is earned. It closes the gap between AI-assisted intent and network reality by providing a continuous loop of safety, validation and learning.

What makes EDA’s AIOps work

EDA’s AIOps capabilities are native to its architecture and enabled by:

• Declarative automation: The AIOps engine parses resources that fully describe intent-based configuration instead of “guessing” the network design from thousands of configurations.
• Abstraction: EDA supports multiple network operating systems (NOSs), including Nokia SR Linux, Nokia SR OS and NOSs from Arista and Cisco.
• Streaming telemetry (gNMI): EDA uses real-time, on-change state data for alarming, intent operational status and drift detection.
• Workflow engine: The AIOps engine uses repeatable, reliable, programmatic tools for tasks where hallucinations are unacceptable.
• Extensibility: The EDA Store lets operators dynamically add new resources and automation applications to extend EDA’s AIOps capabilities.
• Git-backed configuration: EDA enables golden configuration snapshots, CI/CD workflows and precise rollbacks.
• Digital twin validation: A fully integrated, real-time network digital twin lets operators safely simulate changes, plan rollbacks and validate intent before pushing anything to production.
• Natural language engine: The AIOps engine is fine-tuned for networking use cases, with low hallucination risk.

EDA is an ecosystem of apps, agents and analytics services bound by a common event bus and declarative state model. This is what gives it flexibility and power.

Real-time observability with streaming telemetry

EDA’s operational intelligence is enabled by gNMI-based streaming telemetry. Unlike polling-based monitoring that pulls data at intervals, streaming telemetry pushes real-time network state information into EDA’s event pipeline. This gives EDA a continuous, high-resolution view of the fabric’s state.

This real-time awareness fuels EDA’s AIOps capabilities. When a BGP neighbor drops or interface errors spike, EDA informs operators immediately. EDA uses its intent-driven model to constantly compare the observed and desired states. It surfaces drift, triggers alarms and provides clear, actionable insights to guide operators’ next steps.

Why it matters now

According to Uptime Intelligence, human error causes between two-thirds and four-fifths of network outages. In a world of increasing workload mobility, cloud-native applications and AI/ML pipelines, downtime is intolerable.

When done right, AIOps accelerates and de-risks operations. With EDA, Nokia is delivering on the full promise of automation to make networks faster, safer, smarter and more resilient.

See EDA in action

Watch our “EDA in Action” series to see how EDA enables operations teams to move fast with confidence.