Research by Veeam shows that only seven percent of organizations are truly ready for AI, despite nearly nine out of ten already deploying or testing AI agents.
The Veeam Data and AI Trust Gap Report, based on a global survey of 600 senior managers, highlights a growing gap between AI ambitions and the reality within organizations. While AI applications are being picked up quickly, governance, visibility, and control over data often fall short.
The results show that companies primarily struggle with ensuring reliable, secure, and compliant data. Only seven percent of the surveyed organizations meet the requirements to implement AI responsibly and at scale, while 88 percent are already experimenting with AI agents.
Rapid AI Adoption, Low Trust
The study shows that 25 percent of organizations indicate their AI initiatives have been delayed by data challenges. Only 28 percent are confident they can monitor and adjust AI systems when they operate outside predetermined parameters. CEOs are generally more optimistic than technical leaders: 65 percent believe they have a full AI inventory, compared to 48 percent of technical executives.
Furthermore, nearly half of CEOs believe that reliable data can lead to revenue growth of more than 25 percent. At the same time, 83 percent of CEOs feel the pressure to accelerate the expansion of AI and data capabilities, although operational teams do not always share this confidence.
Shadow AI and Evolving Risks
The rise of shadow AI poses an additional risk. 95 percent of respondents report unauthorized AI use within their organization. Only a quarter offer approved alternatives, leading to increased cyber risks. Additionally, 44 percent cite shadow AI as the fastest-growing cyber risk, while European AI legislation is already influencing investment strategies.
The nature of disruptions is changing along with the autonomy of AI systems. Errors occurring at machine speed are harder to detect and require precision recovery. Only a minority of organizations can quickly identify which systems, actions, or data are involved in AI incidents.