Topic starter
11/11/2026 6:01 pm
TL;DR: Telemetry growth is outpacing observability economics, with Apica reporting 54% of enterprises saw data volumes triple and 69% say observability costs now exceed compute and infrastructure combined. Monitoring agentic systems now depends on pipeline design that can absorb extreme scale without making visibility unaffordable.
Editorial analysis by NHI Mgmt Group, based on content published by Apica: “The Infrastructure Debt Agentic AI Is About to Call In”.
By the numbers:
- 54% of enterprises saw their telemetry data volume triple in the past 12 months.
- AI and machine learning workloads now account for approximately 43% of that expansion.
- 69% of agentic AI projects already have observability costs that exceed compute and infrastructure costs combined.
Key questions
Q: Why do autonomous systems drive observability cost so quickly?
A: Autonomous systems drive observability cost quickly because they generate high-frequency logs, traces, and tool-call events from every runtime decision, retry, and failure path.
Q: How does telemetry routing improve agent governance?
A: Telemetry routing improves agent governance by sending the right events to the right control plane, so approvals, tool use, policy violations, and privilege changes are easier to review.
Q: How do teams size telemetry pipelines for agentic AI growth?
A: Teams size telemetry pipelines by estimating event volume, cardinality, and burstiness from agents, then setting ingest, buffering, indexing, and storage limits to absorb peaks.
Practitioner guidance
- Audit telemetry architecture against projected agentic growth Measure current ingestion, routing, storage and search capacity against plausible agentic AI growth scenarios, then identify where the architecture fails first under 6x, 9.5x and extreme-volume conditions.
- Separate evidence tiers from raw telemetry volume Define which agent events must be retained at full fidelity for audit and attribution, which can be sampled, and which can be transformed before storage.
- Treat observability cost as a risk metric Report monitoring spend alongside deployment risk when agent telemetry growth threatens to force reduced retention, delayed monitoring or selective blind spots.
Bottom line: Agentic AI is turning telemetry into a scale and governance problem because the data path itself becomes a limiting control.
Glossary terms: Telemetry pipeline, Agentic AI, Observability, Non-Human Identity (NHI)
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
11/11/2026 6:06 pm
Telemetry debt is becoming governance debt: once agentic workloads make observability unaffordable, organisations start pruning the very evidence they need for control. That creates a blind spot across incident response, audit and service accountability, because visibility is degraded not by attacker action but by architecture choices. The practitioner conclusion is simple: telemetry design now belongs in the same governance conversation as access control and logging policy.
A few things that frame the scale:
- 53% of security leaders expect AI to run major portions of their infrastructure autonomously within the next three years, according to the 2026 Infrastructure Identity Survey.
- 67% of organisations still rely heavily on static credentials despite the risks they pose to agentic AI deployments, according to the 2026 Infrastructure Identity Survey.
A question worth separating out:
Q: What is the difference between sampling and retention in observability?
A: Sampling decides which telemetry events are collected or kept for analysis, usually by dropping a proportion of logs, traces, or spans at ingest or during processing. Retention is how long stored telemetry is preserved before deletion, so it controls historical depth, compliance hold periods, and the amount of searchable evidence available.
👉 Read our full editorial: Agentic AI makes telemetry pipeline architecture the cost factor