Splunk's State of Observability 2026 Report Highlights AI's Double-Edged Sword for IT Operations

Cisco, through its Splunk business unit, 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: the rapid adoption of AI agents is outstripping organizations' ability to effectively monitor them. A significant 83% of respondents expressed concern about the unintended consequences of AI agents, and over half admitted they cannot accurately control AI costs. The report emphasizes that while AI agents are being deployed to fix and prevent issues, their inherent unpredictability necessitates a renewed focus on observability practices to restore stability and financial predictability. This matters deeply to practitioners because the proliferation of AI agents is fundamentally changing the landscape of IT operations. The promise of AI-driven automation is attractive, but the reality is that these agents can introduce new blind spots and cost overruns if not properly managed. For DevOps and SRE teams, this means that traditional observability tools, while still necessary, are no longer sufficient. The report identifies a group of "leaders" who are successfully navigating this new era by focusing on "agent observability" – the ability to monitor AI agents themselves, their behavior, costs, and impact. These leaders are 2.4 times more likely to exceed their observability ROI expectations and report significant annual business value. This trend aligns with broader shifts in cloud and DevOps, where the complexity of distributed systems has long driven the need for robust observability. The introduction of agentic AI adds another layer of abstraction and dynamism, making traditional monitoring approaches even less effective. The report's findings resonate with other industry observations, such as the increasing importance of observability for cost management and the growing adoption of open standards like OpenTelemetry to handle diverse data sources. The emphasis on connecting observability signals to business impact is also a recurring theme, as organizations seek to move beyond simply identifying technical issues to understanding their direct financial and operational consequences. In practice, this means several things for technical professionals. Firstly, there's an urgent need to implement robust "agent observability" solutions that can track the performance, cost, and behavior of AI agents. This includes monitoring model quality, usage, security, and privacy. Secondly, the report suggests that verbose logs, rather than AI infrastructure itself, remain the top driver of observability costs for many organizations. This implies a need for more intelligent data sampling and routing strategies to focus on high-value signals and reduce unnecessary data ingestion. Finally, the increased collaboration between ITOps, engineering, and security teams, driven by the need to monitor AI workloads, underscores the importance of breaking down silos and fostering a more unified approach to operational governance. Practitioners should actively seek out and implement tools and processes that facilitate this cross-functional collaboration to ensure the safe, cost-effective, and predictable operation of AI-driven systems.