A few months ago, we first introduced Operations Agent in public preview. This tool, one of Monte Carlo’s in-platform AI agents, is a powerful partner designed to help data teams move from detection to resolution faster.

Today, we’re excited to share that Operations Agent is now generally available, with added capabilities that unlock new layers of efficiency for detection, triage, and resolution workflows. 

From the beginning, our vision for Operations Agent was simple: data and AI teams don’t need another dashboard. What will really elevate their day-to-day workflows is an accelerant that can take direct action on observability tasks because it deeply understands their data, their patterns, their infrastructure, and their environment.

Operations Agent does exactly this. You can ask it to create a monitor for a critical table, or have it dig into why a monitor is breaching constantly. You can quickly identify misconfigured monitors, missing audiences, or weak thresholds with just a simple, natural-language conversation. 

What’s particularly compelling about Operations Agent is that it doesn’t just advise or make recommendations. It takes action, so that your data and AI engineers can step away from a many-step manual process and turn their attention to more creative, complex problem-solving. 

We’ve already seen Operations Agent deliver this kind of value to many of our enterprise customers, and we can’t wait to see how it continues to transform workflows. Read on to learn more about its new capabilities and the use cases these support. 

What’s new with Operations Agent

New one-click draft monitors

In its initial release, Operations Agent generated YAML whenever it recommended a monitor deployment, which required copy and pasting the configuration and was most suitable for engineers.

 Now, when the agent recommends a monitor, any member of the data team can create a draft with a single click, no copy and pasting or parsing required. The monitor is generated directly in Monte Carlo, where you can review, edit, comment, and collaborate before publishing it live.

This eliminates technical overhead for non-technical users and democratizes monitoring. It also reduces context switching across all types of users, speeding up a critical part of using Monte Carlo. 

For teams scaling observability across dozens or hundreds of data assets, this reduction in friction compounds quickly.

A seamless transition from Alerts to Operations Agent

Alerts are moments of urgency. The faster you understand what happened, the faster you can resolve it. 

That’s why we have better integrated alerts in Monte Carlo with Operations Agent. Not only do alerts now open with a concise summary that captures the key details of the issue – but users can now prompt Operations Agent for more context into the alert and to help with next steps. 

This is all with just the single click of a button, streamlining the experience of going from an alert to remediation. 

Deeper integration with Troubleshooting Agent

Even more powerful is the new handoff between the Troubleshooting Agent and Operations Agent. After Troubleshooting Agent completes its analysis of a problem, follow-up questions seamlessly transition into the Operations Agent side panel.

This preserves continuity and investigative context around an issue. For data and AI teams under pressure to reduce mean time to resolution, this continuity is crucial. It keeps investigations structured, contextual, and efficient, especially when multiple stakeholders are involved.

Built in Monte Carlo Support

Another exciting upgrade to Operations Agent is its built-in Monte Carlo support. 

The agent accelerates onboarding for new users by surfacing deep insights on how the Monte Carlo platform works, offering instructions and tips on how to create validations, set up alert conditions, define custom metrics, and engage with all other product capabilities. 

Additionally, Operations Agent enables you to more easily ask humans for help by connecting directly to Monte Carlo Support. 

Users can request assistance, or the agent may proactively suggest it if it’s dealing with a complex issue that requires more intervention to resolve.

When that happens, the system generates a detailed summary of the entire interaction: what was investigated, what was attempted, and what remains unresolved. You can edit it, submit it as-is, or cancel the request. Once created, the ticket is handled like any other support case.

For teams, this eliminates a frustrating aspect of resolving issues with support: rewriting context from scratch. The investigative history carries forward intact, ensuring faster, more informed engagement. It’s a seamless bridge between AI assistance and human expertise.

Unlocking new use cases

Operations Agent has already been unlocking new ways of working for Monte Carlo customers. With these added capabilities and its full GA debut, the path is paved for upleveling data team efficiency in ways we are continuing to discover every day. 

Teams can now run regular monitoring health audits using Operations Agent instead of relying on tribal knowledge. They can democratize monitor creation, allowing more stakeholders to participate without increasing operational complexity. They can move from alert, to insight, to action inside a single, AI-guided experience.

AI teams, in particular, gain a stronger safety net. Data quality issues that could previously cascade into model degradation, for example, can now be identified, investigated, and escalated. And these actions can all be done through a unified interface.

Let’s take a look at some specific use cases that our newest Operations Agent capabilities empower end users to pursue. 

Proliferating insight across non-technical users

Non-technical users, such as analysts or product managers, are often key stakeholders of observability data, and yet they almost always face barriers to entry with observability products. 

They need insight into what’s happening with the upstream data they rely on for business decisions, but they can’t self-serve because the tools that house that data require engineering expertise. This gatekeeping of information serves no one; in fact, it largely hinders organizations from making real data-driven decisions at the speed they need to in order to stay competitive. 

With Operations Agent’s new, in-product monitor draft creation, non-technical users can easily propose and iterate on monitors without touching YAML.

This democratizes access to insight, lowers the barrier to participation, and expands data reliability ownership beyond a small technical group.

Faster incident response for AI & data pipelines

Mean time to resolution is consistently a metric that data + AI teams aim to improve, as incidents and downtime are costly. With the exponential growth of data stemming from the rise of AI and agentic systems, the number of incidents only multiples – and the need for faster incident response becomes critical for survival. 

The seamless agent handoff between Troubleshooting Agent and Operations Agent bridges the gap between incident root cause analysis and active remediation: 

• Analysts and engineers can quickly understand impact and direct next steps 
• AI teams can trace upstream data issues affecting models 
• Cross-functional teams can collaborate immediately without losing context

This is especially critical in AI-driven environments where a myriad of data issues – freshness, schema changes, code changes, system disruptions, etc. – create more incidents than teams can realistically keep up with. Accelerating many of these activities through agent-to-agent relationships lets teams handle scale effectively, reducing incident response time and preserving business-critical functions. 

Monitoring hygiene and governance at scale

One of the biggest challenges in scaling a data reliability program isn’t creating monitors, but rather maintaining them.

As data environments evolve, monitors almost inevitably drift. Some start breaching constantly and become background noise. Others lose assigned ownership as teams change. A few become subtly misconfigured, weakening their effectiveness. Over time, this doesn’t cause a dramatic failure, but it still chips away at trust within the organization and even beyond. 

Operations Agent, particularly with its new capabilities, lets teams proactively audit their monitor health instead of waiting for complaints about noisy alerts or missed issues. 

In seconds, the agent can surface monitors breaching, for example, more than 90% of the time and suggest tighter thresholds. Or, it can identify monitors with no assigned audience and recommend clear ownership. It can also flag misconfigurations and propose fixes.

What used to require manual reviews and deep institutional knowledge becomes an on-demand health check that even new team members can execute. 

This keeps observability strong as the organization grows, and, for AI teams, it ensures that upstream data issues don’t quietly degrade model performance. 

The future of AI-powered data operations is now

Modern data and AI systems have become too complex for manual workflows alone – anyone working in the space has come to this realization. Teams need automation, but not blind automation. They need intelligently driven gears that fit into the larger mechanics of their data ecosystem, each agent humming along harmoniously with the others.

Operations Agent represents that shift in how we work. As we celebrate the GA of this observability workhorse, Operations Agent is ready to support organizations at scale, helping data and AI teams build more resilient pipelines, reduce operational drag, and move faster with confidence.

If you’d like to see it in action, we’re hosting an upcoming live product demo where we’ll walk through real-world workflows and show how teams can use Operations Agent to solve real data and AI problems.

You can also explore the documentation to dive deeper into its capabilities. 

If you’re already finding value in Operations Agent, we’d love to hear about it! You can reach out to us directly to share how this agent is empowering your teams, and we’re excited to see what you build with it next.

Our promise: we will show you the product.