Agentic AI telemetry is piling up faster than most enterprises can pay to monitor it. In a new TechRadar Pro Perspectives piece, Apica CPTO Andi Mann argues the infrastructure under AI agents needs as much attention as the models themselves.
Agentic AI telemetry by the numbers
Andi draws on Omdia research commissioned by Apica, which surveyed more than 300 IT decision-makers across North America and Western Europe.
- Telemetry volume already grew an average of 3.7x in the past 12 months.
- Respondents expect agentic AI workloads to push that to 9.5x within 2 years.
- 69% of teams running agentic AI now spend more on observability than on compute and infrastructure.
- 59% have killed or delayed at least one AI deployment because monitoring costs ran too high.
Why the fix is a telemetry pipeline
Andi’s answer is architectural. Put an intelligent layer between your telemetry sources and your observability backends, and filter and route data before it lands somewhere expensive. Omdia found teams with a telemetry pipeline in place are 50% more likely to be ready for 6x or greater data growth.
The full article also covers why fraud detection and compliance agents are often the first to get shut off, and why buyers are more open to switching vendors than you’d expect. If you’re already feeling the squeeze, see how Apica cuts observability spend.
