AI is in production at 97% of organizations, but controls lag behind
New BARC research finds widespread operational use of AI alongside persistent gaps in governance
BARC has published the global topical survey “The State of Data and AI Governance,” based on responses from 234 data, AI, IT, and business leaders.
AI has moved firmly into daily operations. Organizations deploy it across document analysis, data management and integration, analytics, software development, security and privacy, customer service, and other business processes. In total, 97% of respondents have at least one AI use case in production.
Which of the following AI use cases has your organization deployed? (n=234)
Governance maturity has not kept pace with this broad adoption. Two thirds of surveyed organizations have experienced an AI-related incident in the past 12 months, including privacy breaches, harmful or incorrect outputs, and unauthorized use. The consequences reach customer trust, regulatory exposure, and financial performance.
The study also reveals a gap between established data governance and AI-specific controls. Many organizations have tools for data quality, access, and metadata, while capabilities for monitoring models and governing agents remain less common. Only 26% use AI model monitoring and observability, and 10% have implemented agent governance.
“AI adoption has moved faster than governance,” says Kevin Petrie, VP of Research at BARC US and co-author of the study. “Organizations now need enforceable controls that connect data, models, agents, applications, and infrastructure. This creates a sound basis for responsible AI use in production.”
There are signs of progress. BARC classifies 29% of respondents as AI Leaders, up from 20% in early 2024. These organizations report fewer incidents in most categories and apply AI-specific controls more consistently.
The study examines governance structures, responsibilities, technology, architecture, data quality, sovereignty, and agent controls. It provides recommendations for turning policies into operational safeguards without slowing productive AI use.
Access the study
“The State of Data and AI Governance” is available for free download thanks to the support of Boomi, Board and SoftProject.
