Veeam SVP: Poor Data Quality Sinks 80-90% of AI Projects
Shiva Pillay, SVP at Veeam Americas, stated that 80-90% of enterprise AI projects fail due to fundamental data issues. He noted that 70-90% of enterprise data is unstructured and was never designed for AI, making modernization a critical prerequisite. Pillay highlighted the lack of consensus on who owns the data strategy—CDO, CIO, or CISO—as a major organizational hurdle.
- The financial impact of poor data quality is substantial, costing organizations an average of $12.9 million annually, with some estimates placing the cost for U.S. businesses as high as $3.1 trillion per year. A single incident at Unity Software, where bad data was ingested, resulted in a $110 million revenue loss. - Recent studies reinforce the high failure rates, with a 2025 MIT analysis finding that 95% of generative AI pilots fail to deliver measurable business returns, despite billions in enterprise investment. This is often attributed to a failure to address organizational and operational readiness, rather than technological shortcomings. - The challenge with unstructured data is its sheer volume; it accounts for an estimated 80-90% of all data generated today. Projects involving this type of data, which requires specialized expertise in NLP or computer vision, can take 2-3 times longer than those using structured data. - In response to these data challenges, the role of the Chief Data Officer (CDO) has evolved from a defensive focus on governance and compliance to a strategic business leader. A 2024 Deloitte survey found that 72% of CDOs now report directly to the C-Suite, emphasizing their central role in aligning data initiatives with business outcomes. - Beyond data quality, significant barriers to AI adoption include a shortage of skilled talent and difficulties integrating AI with legacy systems. A survey of organizations with mature AI implementations found that 34.5% cite a lack of AI infrastructure skills as their primary obstacle. - As the industry moves toward more advanced agentic AI, Gartner predicts that over 40% of these projects will be canceled by 2027 due to rising costs, governance challenges, and an inability to prove clear ROI. Forrester similarly predicts that three out of four firms attempting to build advanced agentic architectures independently will fail. - Cost management has rapidly become a critical success metric for AI projects, jumping from the 7th most important factor in 2024 to the 2nd in 2025. A recent study found 51% of organizations reported that their software costs for AI exceeded expectations. - The "garbage in, garbage out" problem creates a significant time sink for technical teams, with some studies indicating that employees can waste up to 50% of their time correcting and verifying data issues instead of focusing on innovation.
