AWS launched CloudWatch Omni for monitoring the quality of AI agents

Company Amazon Web Services introduced the CloudWatch Omni service, designed to monitor the operation of autonomous systems in production. The new tool shifts the focus of observability from the classic check of "broken or not" to evaluating the logical correctness of task execution.

The platform analyzes the quality of AI agents' work at the level of individual runs. The system tracks the correctness of tool selection, the logic of decisions made, and causal relationships between model steps. This allows developers to identify implicit logical errors, when the system formally works without failures, but produces an incorrect result.

The need for such tools arose due to companies' transition to complex multi-agent architectures. Traditional metrics like response latency and HTTP error percentage do not cover the specifics of generative software, where a failure can be an incorrect sequence of API calls.

The service is already available for integration into existing AWS infrastructures. The tool helps teams scale AI production deployments without the risk of losing control over the behavior of autonomous agents.