Nune Isabekyan, co-founder and CTO of OpsWorker, will speak at WeAreDevelopers World Congress in Berlin, July 8-10, 2026.

WeAreDevelopers World Congress is Europe's largest developer conference - 15,000 developers, 500+ speakers, 8,000 companies gathering in Berlin every year to discuss where software engineering is actually heading. This July, OpsWorker's co-founder and CTO Nune Isabekyan will be on stage with a talk that goes directly at one of the central tensions in the AI tooling space right now.

The Talk: Beyond Chat

The session is titled "Beyond Chat: AI Workflows That Actually Investigate Alerts (So You Don't Have To Know Everything)."

The premise is simple but has significant implications for how teams think about AI tooling for operations: chat-based AI is the wrong interface for incident investigation. Not because AI cannot help - it can, and significantly - but because the chat paradigm fundamentally misunderstands where the cognitive load actually sits during an incident.

"You build it, you run it" sounds great until you're on-call for a Kafka consumer lag spike at 3 AM - and you spent the last six months building the React frontend, not the event pipeline.

Modern teams own their services end-to-end, but no one can be an expert in everything. When something breaks at 3 AM, the problem is not that engineers lack access to AI. The problem is that the chat interface asks them to provide context they do not have under pressure - which service to look at, which logs to pull, what the relevant timeframe is. It assumes the engineer already knows where to start. That is exactly what they do not know.

Why the Chat Paradigm Fails

Nune will break down the specific failure modes of chat-based AI for incident investigation:

  • It expects you to provide context you don't have at 3 AM
  • It assumes you know where to look - the exact thing you're trying to figure out
  • It burns time with back-and-forth when every minute of an incident has a cost
  • It interrogates instead of investigates

The chatbot asking "Can you tell me more about your cluster configuration?" is not a useful response when the pager has been screaming for four minutes.

What Actually Works

The second half of the talk shows what investigation looks like when the AI workflow starts automatically - not when you ask, but when the alert fires.

This means automatically discovering affected Kubernetes resources, correlating metrics with recent deployments, querying the right logs without being prompted, and delivering a root cause hypothesis with supporting evidence. The engineering challenges behind making this work reliably - orchestrating tools across Kubernetes, logs, and metrics; solving the "where do I even start?" problem; building outputs that explain unfamiliar systems - are the substance of the engineering portion of the talk.

Nune will also be direct about what breaks when you let AI loose on production - the failure modes, the edge cases, the places where the system behaves unexpectedly - because understanding the limits is as important as understanding the capability.

Live Demo

The talk closes with a live demo comparing two approaches to the same simulated 3 AM alert:

  • The chatbot experience: back-and-forth, context requests, time burned
  • The workflow experience: "Your deployment rolled out 47 minutes ago with a memory limit reduction. Three pods are OOMKilling. Here's the diff and the kubectl command to rollback."

The contrast is deliberately stark. Not because the chatbot represents a bad implementation, but because it represents a different - and less appropriate - paradigm for the problem.

WeAreDevelopers World Congress — Berlin, July 8-10, 2026

WeAreDevelopers World Congress brings together the global developer community at CityCube Messe Berlin for three days of talks, workshops, and the largest tech expo in Europe. The main program runs July 9-10.

If you are attending and want to connect with the OpsWorker team, reach out at [email protected] or find us at the event.

Session link: wearedevelopers.com/world-congress/agenda/sessions/beyond-chat-ai-workflows-that-actually-investigate-alerts-so-you-don-t-have-to-k-1141382

Learn more about OpsWorker: opsworker.ai

Tagged in:

Kubernetes AI SRE Agent Incident Management Agentic Ops Weekly