Monte Carlo announced the launch of observability agents, a suite of AI agents built to accelerate monitoring and troubleshooting workflows to improve data + AI reliability.

Monte Carlo’s Monitoring Agent recommends data quality monitoring rules and thresholds, which can then be deployed with the push of a button. The Troubleshooting Agent investigates, verifies, and explains the root cause of specific data quality issues while also providing the recommended next steps for resolving them.

Both agents are are not just making simplistic recommendations based on data profiles, but leveraging a sophisticated network of LLMs, native integrations and subagents to gain full visibility into the data estate across data, systems, transformation code, and model outputs.

“AI agents are only as powerful as they are informed,” said Lior Gavish, co-founder and CTO, Monte Carlo. “Our AI agents can execute more sophisticated analyses that are truly useful because they are reviewing data samples to determine what the data looks like, metadata to understand the larger contextual meaning, and query logs to understand how the data is used.”

Monte Carlo’s Monitoring Agent, now generally available, identifies sophisticated patterns and relationships across a dataset that would otherwise be missed by more traditional profiling methods.

For example, the Monitoring Agent may identify that a product SKU id field always starts with “950” followed by 4 unique digits for certain product categories and not others, or that a certain product SKU always has a higher order amount than another. It uses context on how fields are used to prioritize and rank the most critical alerts – providing the most coverage with the least amount of noise.

It then automatically generates a monitor that can be easily understood and deployed across all members of the data team. To date, the Monitoring Agent has made thousands of monitor recommendations with an impressive 60% acceptance rate.

Gavish conservatively estimates the Monitoring Agent increases monitoring deployment efficiency by 30 percent or more.

Monte Carlo’s Troubleshooting Agent, with a general release scheduled for Q2 2025, investigates, verifies, and explains the root cause of specific data + AI quality issues.

The agent tests hundreds of different hypotheses across all relevant tables within a dataset to understand if the root cause of a specific issue is a result of receiving bad data from the source, an ETL system failure, a transformation code mistake, or incorrect model output.

This process leverages dozens of subagents investigating in parallel and takes only a couple of minutes to complete. As a result, data teams can reduce the average time to resolve an incident by 80 percent or more.

Monte Carlo continues to develop and deploy AI agents with a security first mindset. Customer data is never stored by Monte Carlo nor used to train AI models. Only users with the appropriate roles and permissions can access data samples, which can be disabled entirely if desired.

The observability agents automate powerful monitoring and resolution tasks, but never directly manipulate, change, or act upon your critical data and key systems (read-only). This ensures they don’t create more reliability issues than they help resolve.

Monte Carlo plans to extend the capabilities of these agents in the next year to further accelerate detection and resolution of reliability issues, providing end-to-end observability across the data + AI lifecycle. This represents a fundamental leap in Monte Carlo's strategic evolution from a standalone data observability pioneer to a comprehensive data + AI observability solution. By unifying monitoring and troubleshooting across both data pipelines and AI systems, Monte Carlo is addressing the critical need for consistent reliability standards as organizations increasingly invest in technologies to power their AI applications.