New research commissioned by Red Hat suggests that the rapid adoption of artificial intelligence across UK organisations is outstripping the systems designed to control it, raising questions about how enterprises will manage risk as AI becomes embedded in core operations.
The study, conducted by Censuswide across 500 IT leaders in the UK, France, Germany and Italy, found that 75 per cent of UK respondents lack robust AI governance frameworks. This sits alongside high levels of adoption, with 87 per cent of business IT leaders reporting that their organisations are already using agentic AI systems.
The gap between deployment and oversight points to a wider structural issue. As AI systems move beyond experimentation into autonomous decision-making, governance is no longer a peripheral concern but a central requirement for maintaining control over data, infrastructure and outcomes.
Governance lags behind adoption
The findings highlight a growing disconnect between the pace of AI implementation and the ability of organisations to manage it effectively. While companies are increasingly deploying agentic systems capable of acting independently, only a quarter of respondents said they have strong governance frameworks in place.
This imbalance is compounded by a lack of visibility across key components of AI deployment. Red Hat warned that many organisations do not have sufficient oversight of their data environments, infrastructure or relationships with AI providers, leaving them exposed as systems scale.
The issue extends to vendor dependency. Although concerns around AI sovereignty and control are rising, only 67 per cent of UK IT leaders reported having an exit strategy in place. For the remaining third, switching providers could present significant challenges, with 43 per cent of respondents indicating that such a move would have a moderate to significant impact on their business.
This suggests that many organisations are embedding AI into their operations without fully understanding the implications of long-term reliance on specific platforms or suppliers.
Control and sovereignty come into focus
The research also points to a growing demand for stronger regulatory frameworks. UK respondents were the most vocal in Europe in calling for policy measures that enforce open source principles, with 89 per cent supporting such regulation compared with lower levels in France and Germany.
At the same time, eight in ten decision makers said companies should have greater control over how AI systems are built and deployed, particularly in a market where many leading providers are based outside Europe. This reflects broader concerns about sovereignty, not only in terms of data location but also in relation to technological dependency and operational control.
Stuart Harvey, chief executive of Datactics, said the issue extends beyond governance frameworks alone. He argued that many organisations lack the underlying data readiness required to support AI deployment effectively, with inconsistent or poorly governed data environments becoming more visible as AI systems scale.
Harvey suggested that governance is often treated as a compliance exercise rather than an integral part of system design. In practice, static frameworks may struggle to keep pace with AI systems that operate continuously and adapt in real time, requiring more dynamic approaches to oversight.
The findings underline a broader tension in the development of artificial intelligence. Organisations are under pressure to adopt new technologies to remain competitive, yet the structures needed to manage those technologies are still evolving.
As AI becomes more deeply embedded in enterprise operations, the consequences of this imbalance are likely to become more pronounced. Without stronger governance, clearer data strategies and greater control over infrastructure, the risk is that organisations will scale systems they do not fully understand or control.
The question for businesses is no longer whether to adopt AI, but whether they can do so in a way that maintains accountability as systems become more autonomous.
