Businesses are rapidly deploying AI agents across their networks, often outpacing the implementation of essential monitoring procedures and security controls. This accelerated integration leaves organizations vulnerable to a variety of security threats and complications commonly linked to AI technology, including “hallucinations,” as discussed in a recent report published by Cisco on October 20, titled Splunk State of Observability 2026.
Despite the eagerness of many organizations to tap into the transformative potential of AI within their operations, the report highlights a significant lack of strategic planning to ensure the safe utilization of these powerful tools. The findings reveal that 38% of the IT operations and engineering respondents lack clearly defined ownership and accountability when it comes to monitoring AI deployment. This gap in responsibility can largely be traced back to the traditional divisions present within the IT department, which encompass several domains including software engineering, infrastructure management, web development, and information security.
The fragmentation of responsibility complicates the monitoring of AI use within organizations. A staggering 42% of survey participants reported that their principal challenge in observing AI arises from the fact that various teams are responsible for its oversight. Mark Woods, the field CTO and chief technical advisor for EMEA at Splunk, encapsulated this predicament succinctly: “When agent observability is everyone’s job, it becomes nobody’s job.” This reflects a deeper organizational issue, suggesting that without clear delineations of responsibility, the integrity of AI systems is at risk.
As businesses grapple with the management of AI agents at a larger scale, they also confront the unintended consequences of rapid AI adoption. While 85% of respondents emphasized the importance of having visibility into AI hallucinations, only 58% felt confident in their organization’s current capability to effectively monitor these phenomena. Identifying when an AI agent is generating erroneous information or acting upon such errors remains a challenging task, with many organizations lacking the necessary tools to prevent such actions.
This oversight could have dire implications for business operations. For instance, a malicious actor might exploit an AI agent’s vulnerabilities to extract sensitive information unlawfully. The absence of appropriate monitoring could allow such nefarious activities to proceed undetected, potentially leading to major data breaches.
Jeetu Patel, President and Chief Product Officer at Cisco, underscored the critical nature of managing these agents at scale: “Managing agents at scale requires reimagining both security and observability. Have our agents been compromised? Are they doing their job well? In the world of agentic AI, these questions are inseparable, and they’re moving at machine speed.” His comments point to the urgent need for organizations to reconsider both their security protocols and their observational strategies to ensure both effectiveness and safety in AI applications.
The report underscores the convergence of observability and security data, emphasizing that collaborative efforts among teams are more important than ever. To effectively respond to and detect issues in real-time, organizations must develop common data platforms, aiding in the deployment of trustworthy agents at scale.
Interestingly, when AI agents are implemented appropriately, they can significantly enhance the capabilities of IT and security teams. The report highlights that 86% of participants reported improvements in the speed of detecting IT or security issues, with 36% categorizing that enhancement as “significant.” These results point to the potential benefits that effective AI deployment can indeed bring to an organization.
However, the findings are not entirely positive. Despite the majority affirming that AI has made them more efficient, a considerable minority—14%—expressed that the introduction of AI agents has complicating issues, making them harder to detect and resolve. A common concern among these respondents is the additional time required to verify the accuracy of the information provided by AI agents, which can overshadow the operational efficiencies gained.
In conclusion, while the integration of AI agents in business environments holds promise for enhancing operational capabilities, it also necessitates a cautious and informed approach. Without proper monitoring, accountability, and strategic oversight, organizations risk exposing themselves to significant vulnerabilities that could ultimately undermine the very efficiencies they seek to realize through AI technology. The path forward requires collaboration, robust security measures, and a commitment to continually evaluating the roles and responsibilities surrounding AI deployment within the IT landscape.