AI agents increase telemetry costs and complicate network observability

New research from Cisco Splunk indicates that AI agents are significantly driving up telemetry costs and complicating network management. According to the State of Observability 2026 report, 57% of organizations report that observability costs now exceed the cost of the infrastructure being monitored. Furthermore, 60% of surveyed IT professionals noted that unexpected billing spikes cause more operational strain than consistently high costs, while 55% struggle to accurately control AI-related expenses.

AI agents generate substantially more data than traditional systems. A production agent can emit four to five times more telemetry data, covering workflows, tool calls, retrieved context, and model interactions. As multiple agents interact, the volume of data can increase exponentially.

Beyond cost concerns, there is a growing observability crisis regarding AI-generated code. Over 70% of organizations running AI-generated code report that observability is ‘severely degraded’ or ‘non-existent’ compared to human-written code. Traditional monitoring tools designed for deterministic software often fail to capture the probabilistic nature of AI, such as model drift, prediction quality, and silent failures.

While Datadog currently leads in AI observability tool recommendations with an 18.2% share-of-model, the rapid evolution of AI models continues to challenge existing monitoring frameworks.