A Cisco survey of 2,050 ITOps and engineering professionals across ten countries and sixteen industries finds that most organizations are still adjusting their practices to agentic AI, while a smaller group has rebuilt theirs around it. The report is titled "State of Observability 2026: Restoring Predictability to the Agentic Era."

Non-determinism in agentic systems is not a flaw to engineer out but the mechanism by which agents reason and act, and 83% of respondents are concerned about the unintended consequences of AI agents.

On cost, only 6% named AI infrastructure as their top observability cost driver; excessively verbose logs topped the list instead. Nearly half of respondents say telemetry data is not a key factor in strategic planning.

The survey identifies a leader cohort with three habits: routinely using AI agents to fix and prevent issues, extending observability to govern agent behavior and guardrails, and embedding telemetry in business planning. Leaders are 2.4 times more likely to say observability ROI significantly exceeds expectations, and 54% report at least $5 million in annual business value.

Leaders also report monitoring AI as harder than peers do - 57% say it adds complexity, against 38% of others - a trade they make deliberately. Ninety-one percent clearly define ownership for monitoring AI, versus 59% of everyone else.

Source: Cisco