Splunk's 2026 Observability Report Highlights AI Agent Unpredictability and the Urgent Need for Cost Control

Cisco, through its Splunk business, has released its sixth annual State of Observability report for 2026, titled "Restoring Predictability in the AI Era." The report, based on a survey of 2,050 ITOps and engineering professionals across ten countries, highlights a growing disconnect: organizations are rapidly deploying AI agents, but their ability to monitor and control these agents' behavior and associated costs is lagging. A key finding is that 83% of respondents are concerned about the unintended consequences of AI agents, and over half (55%) admit they cannot accurately control AI costs. This matters deeply to practitioners because the unchecked proliferation of AI agents can lead to significant operational and financial liabilities. The report indicates that while AI is seen as a driver for innovation, its unpredictable nature in production environments, coupled with opaque cost structures, can undermine its perceived benefits. For engineering and operations teams, this translates into increased troubleshooting complexity, potential service disruptions, and budget overruns. The report also emphasizes that verbose logs, rather than AI infrastructure itself, remain the top driver of observability costs for many organizations. This trend aligns with the broader industry shift towards more autonomous and AI-driven systems, where traditional monitoring approaches often fall short. The increasing adoption of hybrid, multi-cloud, and edge architectures, as noted by Flexera's 2024 State of the Cloud Report, already introduced significant complexity. The integration of AI agents further compounds this, creating dynamic and less deterministic environments. The report suggests that observability is becoming the "north star" for regaining predictability in this agentic era, moving beyond simply collecting data to actively governing AI behavior and connecting telemetry to business value. In practice, this means that practitioners must prioritize "agent observability" – the capability to monitor AI agents and implement guardrails. This includes a focus on understanding AI costs, which are often convoluted across multiple agents and models. The report suggests that leading organizations are addressing this by more aggressively sampling and routing data to focus on high-value signals, thereby optimizing observability spend and achieving better ROI. Furthermore, the need for shared data and increased collaboration between observability and security teams is becoming critical, with 81% reporting increased collaboration due to monitoring AI workloads. This implies a need for integrated platforms and practices that can provide a holistic view of AI agent performance, security, and cost implications, enabling teams to transform AI from an unpredictable expense into a measurable and strategic investment.