Splunk survey exposes widening gaps in AI agent oversight
Businesses are deploying artificial intelligence agents faster than they can monitor their behaviour, leaving gaps in security, accountability and cost control, according to a global Splunk study released by Cisco on Tuesday. The 2026 State of Observability report found that 87% of technology professionals consider AI monitoring a priority, yet 56% cannot measure important workload indicators…
A new survey by Splunk, conducted with Cisco, reveals that companies are rolling out artificial intelligence (AI) agents at a faster pace than they can effectively monitor their behavior. This oversight can create gaps in security, accountability, and cost control. According to the 2026 State of Observability report, 87% of technology professionals view AI monitoring as a priority, yet 56% struggle to measure key workload indicators like the quality of responses generated by large language models.
Additionally, 54% of respondents say their monitoring efforts are not tied to clearly defined business outcomes. The survey, which polled 2,050 IT operations and engineering professionals across 10 countries, highlights challenges as autonomous software transitions from answering questions to performing tasks and interacting with business systems.
Cisco's findings indicate that 83% of respondents are concerned about the potential unintended consequences of AI agents. While 85% consider visibility into fabricated or inaccurate AI outputs essential, only 58% believe their organizations can effectively monitor such "hallucinations." These monitoring weaknesses matter because AI agents can make decisions, execute software commands, and access company information with varying levels of human supervision.
An erroneous response could inadvertently trigger operational actions instead of remaining isolated within a chatbot conversation. Furthermore, responsibility for oversight is fragmented, with engineering, operations, and cybersecurity teams sharing the burden of understanding how agents function, what they access, and whether their actions comply with organizational safeguards.
Cisco's president and chief product officer, Jeetu Patel, emphasizes that managing AI agents at scale necessitates integrating security and observability considerations. Financial concerns also loom large, with 55% of respondents unable to manage AI costs effectively, and 60% finding unexpected billing spikes more disruptive than steady, high spending.
Over half (57%) of respondents believe the cost of collecting and processing monitoring data surpasses the infrastructure costs of the monitored systems. Notably, 67% of respondents currently use a single observability platform to provide a unified view of systems and associated costs. Collaborative efforts between technology and security teams have improved, with 81% reporting that monitoring AI workloads has increased cooperation.
Eighty-three percent of respondents could trace performance issues to security causes, up from 65% in 2025, and 77% now troubleshoot problems jointly with security teams, compared to 62% in 2025. However, differences exist between national markets. UK respondents are more likely to use AI agents to diagnose technical problems (32% vs. 20% globally) and express stronger concerns about unintended consequences (55% vs. global average of 48%).
Saudi Arabia stands out with 98% considering AI enables focus on innovation, compared to a global average of 84%. Cisco classifies organizations with advanced monitoring practices as leaders, with 91% reporting the ability to attribute AI expenditure to specific agents, models, and workloads, compared to 69% of other organizations.
The survey also reveals that 94% of leading organizations collaborate with security teams, compared to 80% elsewhere.
Written by urgent.news from Arabian Post's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
