Splunk survey: AI agents are outrunning the tools to monitor them

A gap between AI ambition and visibility

Cisco released the report, titled Restoring Predictability in the AI Era, on October 20, 2026. It says the urgency to deploy AI agents is "rapidly outpacing the ability to monitor them". Oxford Economics researchers surveyed ITOps and engineering professionals, from practitioners to executives, from February through March 2026.

The numbers point to a gap between intent and execution. According to the report, 87% of respondents confirm monitoring AI is a priority, yet 54% have no clear business outcomes tied to it. In addition, 56% cannot measure AI workload metrics such as LLM response quality. While 85% say visibility into AI hallucinations is important, only 58% claim to be effective at it. Some 83% are concerned about the unintended consequences of AI agents.

Surprise bills matter more than big bills

Cost is the second theme. Splunk argues that the problem is less the size of the bill than its unpredictability. The report finds:

• 55% say they cannot control AI costs, as usage grows and stays hard to predict.
• 60% say unexpected billing spikes cause more operational strain than high baseline costs.
• 57% say their telemetry costs eclipsed the cost of the infrastructure being monitored.
• 67% use a single observability platform, which gives a unified view of where telemetry costs originate.

Kamal Hathi, SVP and GM of Splunk at Cisco, said "the excitement of AI is clashing with the reality of the bill". He argued that putting observability into an AI tokenomics framework can turn AI "from an unpredictable expense into a measurable and strategic long-term business investment".

Observability and security data converge

As agents gain access to the whole tech stack, observability and security teams are working more closely. The report says 81% of respondents see more collaboration between the two since they began monitoring AI workloads. Also, 73% can accurately trace performance issues to security root causes, up from 65% last year. Troubleshooting jointly with security teams rose to 77% from 62% in 2025, and data sharing between teams climbed from 74% to 84%.

Jeetu Patel, President and Chief Product Officer at Cisco, said: "Most companies will soon have far more agents at work than people." He added that questions about whether agents are compromised and whether they do their job well are "inseparable".

What leaders do differently

The research identifies an advanced group of leaders that follow three practices: they rely on AI agents to fix and prevent issues, they closely monitor model quality, cost, usage, security and privacy, and they connect signals to business impact.

• 91% of leaders accurately attribute AI costs, versus 69% of non-leaders.
• Leaders are 2.4 times more likely to say the ROI of their observability platforms significantly exceeds expectations.
• 54% of leaders say observability has yielded at least $5 million in annual business revenue, versus 28% of non-leaders.
• 94% of leaders say monitoring AI increased collaboration with security teams, versus 80% of non-leaders.

Country differences and what it means

Among the 10 countries, India stands out on agent observability: 51% call their practice very effective at monitoring AI model quality, against 35% globally. In the UK, 32% use AI agents all the time to diagnose root cause (20% globally), and 55% strongly worry about unintended consequences. In Saudi Arabia, 98% say AI lets them focus more on innovation, versus 84% globally. In Germany, 74% rate visibility into AI privacy as very important.

For enterprises running agents, the report's message is that monitoring quality, spend and security in one place is becoming a prerequisite, not an add-on. Note that the findings come from a study published by Cisco, which sells Splunk observability products.