groundcover today extended the reach of its observability platform to include a shared workspace where artificial intelligence (AI) agents and human software engineers can collaboratively conduct investigations.

Company CEO Shahar Azulay said the Agent Investigations workspace makes it simpler for DevOps teams to employ AI agents to test a hypothesis when investigating an incident. The overall goal is to provide a more cohesive approach to conducting those investigations versus requiring each engineer to manually summarize their own independent investigations of an incident.

At the core of that capability are two classes of AI agents. An Investigator agent builds a graph of questions and hypotheses, queries logs, metrics, traces and Kubernetes data to test each one and records findings with supporting graphs. A Moderator agent then reviews every proposed conclusion to ensure there is enough evidence to support it. That latter agent, in effect, runs the show by acting as a judge that evaluates the findings of the Investigator agents, said Azulay.

The end result is a shared workspace where AI agents provided by groundcover, or ones that DevOps teams have separately deployed, are, via a Model Context Protocol (MCP) server, able to test a hypothesis against production data, document events that support or refute it and challenge conclusions that lack evidence.

That approach creates a single shared record of the investigation that every engineer and AI agent involved can use to track what questions were already asked, hypotheses tested, evidence collected, and conclusions reached. The overall goal is to bridge the gap between AI agents and humans that require some type of user interface to interact with them, added Azulay. Otherwise, AI agents are simply going to create another type of bottleneck that could be avoided using a common workspace, he noted.

The groundcover platform captures telemetry data from production environments using the OpenTelemetry frameworks and extended Berkeley Packet Filter (eBPF) in a way that doesn’t require IT teams to first instrument every application. The company also recently acquired a Wand platform that makes it possible for IT teams managing Kubernetes clusters to automatically apply the insights surfaced by the observability tools provided by the groundcover observability platform.

In general, spending on observability is on the rise. A recent Futurum Group survey found that nearly two-thirds (64%) of organizations plan to increase observability spending over the next 12 months. As more organizations deploy AI agents across the enterprise, the need for observability platforms to track their behavior is only going to increase to ensure that DevOps teams are able to stay in the loop, said Azulay.

The one thing that is certain is observability platforms themselves will become much more accessible in the agentic AI era. Rather than requiring a DevOps engineer with specialized skills to decipher the insights being surfaced, the average IT administrator can now use natural-language inputs to summarize those insights and generate dashboards as needed to both investigate incidents and, hopefully, proactively prevent them from occurring in the first place.

Frequently Asked Questions

1. What is groundcover's new Agent Investigations capability?
It's a shared investigation workspace that allows DevOps engineers and AI agents to collaboratively analyze incidents using production data. The system maintains a common record of hypotheses, evidence, findings and conclusions, helping teams avoid duplicated investigative work.
2. How does groundcover prevent AI agents from jumping to incorrect conclusions?
Its Investigator agents test hypotheses against logs, metrics, traces and Kubernetes data, while a Moderator agent reviews proposed conclusions to determine whether sufficient supporting evidence exists. This adds an evidence-checking layer, although it does not eliminate the need for human oversight.
3. Why is AI-powered observability becoming increasingly important for DevOps teams?
As organizations deploy more AI agents and distributed applications, understanding system behavior becomes more complex. AI-powered observability can help engineers correlate telemetry, investigate incidents through natural language and identify potential problems without manually navigating multiple monitoring tools.