New BARC research shows AI is in production at 97 percent of organizations, but governance is not keeping pace: two thirds have already experienced AI-related incidents. In EMEA, data quality and data silos hold organizations back far more than in North America.
Ettlingen, Germany, October 28, 2026 – Artificial intelligence is already in production at 97 percent of organizations, but governance is not keeping pace: two thirds reported AI-related incidents in the past twelve months, such as data privacy breaches, biased outputs or shadow AI. These are the findings of “The State of Data and AI Governance: Requirements and Best Practices to Safeguard Innovation”, a study by the analyst firm BARC based on a global survey of 234 data, AI, IT and business leaders. The gap between regions is particularly wide: in North America, 37 percent of organizations qualify as “AI Leaders”, compared with just 17 percent in EMEA.
“AI governance has shifted from theoretical policy work to operational risk management,” write BARC analysts Kevin Petrie and Timm Grosser in the study.
AI incidents are already widespread
Data privacy breaches and biased or harmful outputs each affected 23 percent of respondents in the past twelve months, followed by unauthorized or shadow AI (22 percent) and incorrect AI outputs reaching end users or customers (21 percent). At the same time, many organizations overestimate their own maturity: 29 percent qualify as AI Leaders with mature, executive-led AI programs, up from 20 percent in early 2024, but 60 percent believe their governance is superior to that of their competitors.
Governance tools lag behind AI in production
Most organizations still govern data better than AI models or agents. Data quality management is in use at 58 percent of organizations, while only 26 percent use AI model monitoring, 20 percent use dedicated AI governance platforms and 10 percent use governance tools for agents. To control AI agents, 59 percent rely on human-in-the-loop approval for sensitive actions; only 16 percent have implemented kill switches. “Humans cannot compensate for the lack of automated control systems,” the study concludes.
EMEA works on fundamentals, North America on scaling
The regional comparison shows two different stages of maturity. 79 percent of North American organizations have centralized governance structures, compared with 50 percent in EMEA. EMEA organizations struggle more with the basics: 40 percent name inadequate data quality as a top obstacle (North America: 15 percent) and 35 percent cite siloed data (North America: 17 percent). EMEA respondents also report more shadow AI incidents (29 versus 18 percent), and only 30 percent have formal sovereignty policies for AI processing and storage, compared with 56 percent in North America. “North America has scaling risks and EMEA struggles with foundational risks,” the authors summarize.
[SOFTPROJECT QUOTE – DRAFT, APPROVAL REQUIRED] “The study confirms what we see with our customers across Europe: the AI is ready, but the data isn’t. Governance is not a brake on AI. It is what allows AI to act, with trusted data, clear rules and automated controls that keep up with production,” says [Name], [Title] at SoftProject.
Six recommendations from BARC
BARC derives six recommendations for organizations from the results:
- Treat AI governance as an operating model, not a policy project.
- Close the gap between data governance tools and AI governance capabilities.
- Make human oversight explicit, but do not rely on it as the only control.
- Move governance deeper into the architecture, from the data layer to applications, models and infrastructure.
- Extend data quality to the data AI actually uses, including documents, knowledge bases and retrieval pipelines.
- Measure governance by the risks it reduces.
Availability
The full study is available for free download at https://www.softproject.com/resources-library/ai-governance-study/
Methodology
BARC conducted a quantitative online survey in July and August 2026. After data cleansing, 234 qualified responses from data, AI, IT and business leaders remained for analysis. 61 percent of respondents are based in North America, 37 percent in Europe and 2 percent in the rest of the world. The study was sponsored by SoftProject.
About SoftProject
SoftProject helps organizations turn data into trusted business value. Its platform connects data sources, makes data trustworthy through data governance and data quality, and turns it into governed, automated action, from integration and process automation to AI agents. Organizations in financial services, manufacturing, energy and utilities, the public sector and insurance rely on SoftProject to build a reliable foundation for AI, analytics and digital transformation. More information: www.softproject.com
About BARC
BARC is a leading analyst firm for data and analytics, AI, corporate performance management (CPM) and ESG, with a reputation for unbiased and trusted advice. Its analysts deliver research, events and consulting services for the data and analytics community. BARC has 25 years of experience with data strategy and culture, data architecture, organization and software selection. More information: www.barc.com