Data paradox: poor data quality undermines AI in finance

MindBridge has published research linking widespread data quality problems to delays in finance processes and measurable financial losses, even as organisations expand their use of AI and automation.

The survey of 640 professionals across retail, manufacturing, and energy found that 88.6% experience delays in critical financial workflows because of data issues. It also found that 90% report a direct financial hit from undetected errors, with around 62% describing the impact as moderate to severe.

MindBridge framed the findings as a "data paradox": organisations expect AI to improve accuracy and efficiency, yet poor data quality and weak controls create friction and losses that are hard to detect.

Concerns over risk

The research suggests automation can create new exposure when controls and oversight do not keep pace. Around 40% of businesses said they were somewhat or very concerned that errors, risks, or unusual activity could go unnoticed as they implement AI.

Responses also suggest cutting headcount is not the main driver for adopting AI in finance teams. Only 6% said they view AI primarily as a way to reduce staffing. Many instead cited accuracy, trust, and time savings as the main motivations.

When asked about AI's biggest benefits, improving accuracy and trust ranked highly across all three sectors. In retail, 54% cited that benefit, compared with 45% in energy and 34% in manufacturing. Reducing repetitive manual work also featured prominently: 53% in manufacturing selected that outcome, compared with 48% in energy and 44% in retail.

Sector breakdown

The results varied by industry in both how data problems show up and how severe the impact appears. Retail stood out for the scale of workflow disruption: 94% of retail professionals reported delays caused by data issues, compared with 89% in energy and 83% in manufacturing.

Retail respondents also reported higher anxiety about rapid automation. Nearly 44% of retail leaders said they were concerned that critical risks or unusual activity could go unnoticed as they streamline operations with AI. Budget pressures also appeared more acute in retail, with 43.5% citing budget and resource constraints as the primary barrier to AI adoption, compared with 31% in energy and 28.2% in manufacturing.

In energy, the survey pointed to a mismatch between confidence and day-to-day experience. Some 68.5% said they were confident or very confident in their data for financial decisions, yet 88.6% reported that data quality issues still cause delays. More than half (50.6%) described those delays as moderate to significant.

Energy respondents also reported larger financial impacts than the other two sectors. Some 40% said undetected errors or data quality issues have a major or severe financial impact, compared with 31% in retail and 20% in manufacturing.

Manufacturing respondents reported fewer frequent delays from data issues, with 7.9% saying delays happen often. However, smaller interruptions appear common: 45% reported "some delays", compared with 39% in retail and 38% in energy.

Governance focus

MindBridge positioned the results as a warning to senior leaders that automation-driven speed gains can bring new operational and financial risks if governance and data controls do not keep pace with adoption.

"The 'data paradox' represents a critical friction point for the autonomous enterprise. Our study shows that while teams are racing toward an AI-powered future, they are being held back by data errors and issues that create significant financial and operational drag. Nearly 90% stalled by data quality issues is not a minor friction point. It is a structural gap between the pace of AI adoption and the controls designed to govern it," said Stephen DeWitt, Chief Executive Officer, MindBridge.

DeWitt also highlighted a gap between what leaders believe about the state of their data and what teams experience in day-to-day operations.

"This 'data paradox' is most visible in the disconnect between trust and reality, where leaders are confident and trust their data, but the hard facts show otherwise. Undetected errors are producing real financial damage, at scale, and largely out of sight," said DeWitt.

MindBridge said finance teams and boards need systems that can explain outputs and decisions at a transaction level. It also argued that traditional sampling approaches in finance and audit do not match the volume and speed of automated workflows.

"CFOs, CIOs, and boards need AI systems that show their work and can explain every transaction, data point, or calculation. To achieve this, we need to move away from traditional sampling of financial data towards explainable AI that continuously processes 100% of transactions. Finance is becoming autonomous, but automation without governance is a risk. True digital transformation isn't just about speed; it requires accountability at scale," said DeWitt.