When AI Runs On-Call: How Komodor Is Automating the Toughest Parts of SRE

It’s a scenario every engineer has likely lived through and dreads more than anything else. It’s 2 a.m., and you receive an alert on your phone. Meanwhile, back at the office, dashboards are flashing red, and you need to begin the laborious task of figuring out if it’s real or just noise. This involves manually correlating logs, metrics, Kubernetes events, and infrastructure changes before any kind of fix can even be considered.  Komodor believes AI can do most of that work.

The Kubernetes operations platform has announced Reliability Agents, autonomous AI agents that are built to help teams with incident investigation, root cause analysis, and even execution of approved remediation. For SREs, Reliability Agents could significantly cut down how long they have to spend triaging issues. This shift marks a significant evolution from merely having a chatbot or an assistant.

From monitoring to autonomous investigation

While traditional observability platforms will tell your team that something’s wrong, figuring out what and why is a different story altogether. It’s a manual process that typically requires opening up multiple dashboards and disconnected tools. That’s the hole Reliability Agents aim to plug. They do the heavy lifting of correlating deployments, infrastructure changes, logs, metrics, and runtime data to investigate incidents automatically.

The best part is that, as opposed to simply generating alerts, they provide you with contextual explanations of what happened, why it happened, and what systems were impacted. SRE teams managing complex Kubernetes environments may find that their initial triage time is cut dramatically. 

AI that executes fixes

What’s truly interesting about this announcement is that the platform moves beyond simply providing AI-generated recommendations.

Reliability Agents perform real actions, like restarting workloads, rolling back deployments, or even running predefined operational playbooks, given they’re approved by organizational rules and policies, of course. The key here is that they won’t replace SREs, but instead will help with certain jobs, specifically routine, repetitive operational tasks. Komodor’s stance is that these autonomous agents will help reduce alerts and mean time to resolution without giving up complete control of the incident.

As AI agents are rapidly finding their way into enterprise operations, for SREs, it will mean more than just a fancy chat interface. These agents can investigate and work through a workflow to resolve an issue or even interact with the infrastructure directly to perform real tasks.

Reflecting a larger shift in SRE

For SRE teams, this represents a significant evolution. As Kubernetes environments get more complex, there’s always a lot more data that needs to be analyzed to resolve an incident. Platforms like Komodor’s will help teams by offloading the repetitive work of investigation, allowing engineers to focus on more impactful changes and high-priority work.

Komodor’s announcement shows how platform vendors are beginning to view autonomous agents as tools that make humans more effective as opposed to tools that replace them. Using continuous analysis coupled with approved and controlled automation helps teams improve their reliability without removing human oversight from the incident response process.

While it’s still the early stages for autonomous SREs, this announcement gives us a look at where we may be headed. With more AI agents becoming embedded into our infrastructure, the future of on-call engineering may involve fewer late-night investigations and more intelligent systems that ensure they get a good night’s sleep.