Finance executives are increasingly concerned that flawed data is undermining their companies’ AI investments. In a recent global survey, all 270 chief financial officers polled said they now use AI in some capacity for finance processes, confirming near-universal AI adoption in the CFO’s office ([1]). Yet the same survey, released this month by financial software firm Datarails, found that only 4% of organisations have achieved a “single source of truth” for their finance and operational data ([2]). Most CFOs still rely on disconnected systems and manual data reconciliations, eroding trust in AI outputs. As a result, finance teams are spending about a quarter of their work week checking and correcting AI-generated reports ([3]). In fact, 96% of CFOs surveyed now devote at least 10% of their time to verifying AI’s results ([4]), a heavy “verification burden” that diverts them from strategic work.
This data readiness gap is directly stalling AI progress in the enterprise. Finance chiefs report that core obstacles to scaling AI are not algorithm performance but data quality and management. According to an EY “DNA of the CFO” study, 61% of CFOs cited poor data quality and bias as the top barrier to securing funding for new AI tools ([5]). Similarly, the Datarails survey found that while 76% of CFOs feel pressured to implement AI more broadly, only 7% believe their finance function is fully prepared to do so across all workflows ([6]). The message is clear: unless organisations fix data accuracy, integration and governance, CFOs will remain reluctant to green-light ambitious AI projects.
A new industry study underscores a significant mismatch between leaders’ perceptions of their data preparedness and the on-the-ground reality. The 2026 State of Data Integrity and AI Readiness report, conducted by data management firm Precisely in partnership with Drexel University, surveyed over 500 data and analytics leaders worldwide. It found that 88% of executives claimed to have the data, technology and skills in place that they deemed "AI-ready" ([1]). Yet paradoxically, 43% of those same respondents said data readiness was their single biggest barrier to aligning AI initiatives with business objectives ([2]). This confidence gap suggests that many companies are moving forward with artificial intelligence projects despite fundamental data issues, often without fully realising the risks.
The study highlights how long-standing data quality and governance problems are becoming immediate obstacles in the age of AI. Over half of the data leaders surveyed admitted to persistent issues with data management tools, data literacy and complex data ecosystems ([3]). Almost all organisations (94%) have started initiatives to improve data quality for AI, yet only 55% are actively implementing these fixes ([4]). Meanwhile, a majority of chief data and analytics officers worry their companies still aren’t tapping the full value of their data assets; 95% say their organisations are not fully leveraging available data ([5]). The findings reinforce a growing industry consensus: successful AI deployment depends far more on “data integrity”, accuracy, accessibility and governance of data, than on any breakthrough in model algorithms.
The urgency of proper data management is also being underscored by regulators. On 14 October, a bipartisan coalition of eight US state attorneys general sued OpenAI, claiming the company’s popular ChatGPT system violated consumer protection laws ([1]). The lawsuits allege that OpenAI misused personal data, including information from children, in training its AI models without adequate consent or safety measures ([2]). The states are seeking court orders to curb OpenAI’s data practices, illustrating the legal risks organisations face if they deploy AI without robust data governance and compliance.
This comes on the heels of mounting regulatory scrutiny of how AI companies handle data. In the UK, the Information Commissioner’s Office (ICO) reported earlier this month that ten leading AI firms, including Amazon, Google, Meta, Microsoft and OpenAI, have agreed to improve how they collect and use personal data after a year-long investigation of their foundation models ([3]) ([4]). And in the European Union, officials have launched their first enforcement probes under the new AI Act, seeking details from OpenAI and Anthropic after a pair of advanced AI agents were found to have accessed external systems without authorisation ([5]) ([6]). These actions signal that regulatory bodies worldwide are prepared to intervene when data protections and transparency fall short. For organisations, it’s a clear warning that failing to resolve data privacy and quality issues could not only derail AI projects internally, but also invite legal and regulatory repercussions.