TLDR: OpenTelemetry, the open-source project for telemetry data collection, has introduced semantic extensions to support the observability of artificial intelligence (AI) agents. This development, highlighted by Morgan McLean, a co-founder of OpenTelemetry and a director at Splunk (a Cisco unit), aims to integrate AI agent telemetry with existing observability platforms, eliminating the need for separate monitoring solutions. Cisco is also actively developing AI agents for its Splunk Observability platform, leveraging OpenTelemetry to automate data collection, issue detection, root cause analysis, and remediation.
In a significant advancement for the observability landscape, the OpenTelemetry project has rolled out new semantic extensions designed to enable comprehensive monitoring of artificial intelligence (AI) agents. This initiative, detailed on September 12, 2025, by DevOps.com, marks a crucial step towards unifying the observability of AI-driven systems with traditional application and infrastructure monitoring.
Morgan McLean, a co-founder of the OpenTelemetry project and a director of product management at Splunk, a unit of Cisco, emphasized that these extensions will empower observability platforms to collect and analyze telemetry data generated by AI agents alongside other applications. The overarching goal is to streamline operations by removing the necessity for deploying distinct platforms solely for observing AI agents.
The move comes as the adoption of OpenTelemetry continues to accelerate, driven by its simplified deployment and configuration. OpenTelemetry collectors are now capable of automatically discovering elements within an IT environment from which telemetry data can be gathered, significantly reducing the manual effort previously required for instrumentation.
Cisco is at the forefront of integrating these advancements. Earlier this week, the company previewed a series of AI agents for its Splunk Observability platform. These agents are designed to automate the collection of telemetry data using OpenTelemetry software, detect issues, identify root causes, and apply fixes. Kamal Hathi, senior vice president and general manager for the Splunk business unit at Cisco, noted that the rise of AI agents is set to accelerate the convergence of IT monitoring, observability, and analytics, aiming to dismantle existing silos that hinder agile responses to IT incidents.
Beyond just collection, OpenTelemetry is also extending its reach into ML Ops and AI pipeline tracing. This includes the ability to trace model training jobs in frameworks like TensorFlow and PyTorch, track GPU usage and memory bottlenecks, and export telemetry to specialized tools for profiling. By 2025, observability is evolving beyond traditional metrics, logs, and traces to incorporate AI-driven analysis and automated remediation, with OpenTelemetry serving as the de facto standard for instrumentation.
This integration of AI with OpenTelemetry is expected to fill previously unbridgeable observability gaps, offering capabilities such as adaptive sampling based on system load, automatic context propagation, AI-enhanced data collection, and resource attribute automation. The ability of AI agents to manage the exponentially increasing volume of telemetry data at unprecedented scales will be critical for IT teams.
Also Read:
- Visual Studio Code 1.104 Enhances AI Model Management and Agent Security Features
- Datacom Pioneers AI-Led Modernization, Automating Up to 70% of Legacy Application Code
In related developments, Cisco also launched Cisco Data Fabric, a platform designed to aggregate and analyze machine data at scale without requiring ingestion into a specific platform. This platform will soon be enhanced to support time-series data analysis using multiple AI toolkits. Furthermore, Cisco has revealed plans to build a large language model (LLM) trained on telemetry data, forming the core of its AI Canvas platform, which aims to unify network, cloud, security, and observability data for AI-driven automation of IT tasks.
