Only 9% of organisations say all their data is available and usable for AI, according to a survey of over 1,500 IT leaders from Cloudera.

However, 38% reported that most of their data is accessible.

While the majority (86%) of organisations say they are data driven, many still face technical limitations with the data architecture when it comes to supporting AI workloads. The biggest challenges cited were data integration (37%), storage performance (17%), and computer power (17%), lack of automation (17%) and latency (12%).

Separately, 61% of respondents said siloed data had at least sometimes negatively impacted their ability to scale AI initiatives.

While more than three-quarters (77%) of respondents were confident in their organisation’s ability to secure data used in AI systems, there are still lingering concerns.

Data leakage during model training (50%) was the top concern when it came to AI security, followed closely by unauthorised data access (48%) and unsecure third-party AI tools (43%).

Organisations also cited a lack of visibility or explainability in model outputs (39%) and model manipulation or poisoning (35%) as other top concerns.

Despite these challenges the speed of AI adoption shows no sign of slowing down with 21% saying that AI is already fully integrated into their business processes.

Recommended reading

• A Quarter of UK Firms Have Fallen Victim to AI Data Poisoning
• Is Data Infrastructure Getting Side Lined in the Push for AI in Health?
• Only 2% of Firms Ready for Agentic AI Risks, New Report Warns

A further 54% said AI was significantly integrated and 21% said it was somewhat integrated. Overall, 70% of respondents said they have already achieved significant success with AI initiatives.

“In the last 12 months, AI has shifted from a strategic priority to an urgent mandate, actively reshaping operations and redefining the rules of competition,” said Sergio Gago, Chief Technology Officer at Cloudera.

“But our survey shows that challenges around security, compliance, and data utilisation still remain. Organisations need access to all of their data, wherever it resides and in any form, to govern it securely and unlock real-time and predictive insights.

“As a result, hybrid data architectures are becoming essential, giving organisations the flexibility to manage AI seamlessly across both cloud and on-premises environments.”