Why AI Projects Stumble Before They Begin in Enterprises
George Kurian, CEO of NetApp, emphasizes that most AI failures originate from poor data readiness rather than insufficient infrastructure. He observes that companies often prioritize expensive GPU upgrades and advanced computing resources while neglecting the foundational quality of their data. This misalignment can lead to stalled projects and unrealistic expectations about AI outcomes.
Kurian explains that eighty-five percent of AI project time is devoted to locating, cleaning, and organizing datasets before any model work occurs. Organizations often underestimate the complexity of structuring data across multiple departments and legacy systems, which slows progress considerably. Without a clear data strategy, even the most advanced AI models fail to deliver meaningful results.
The scale of AI adoption in enterprises has grown rapidly, with firms across finance, healthcare, manufacturing, and technology investing heavily. Despite this growth, many organizations continue to treat AI as primarily an infrastructure challenge rather than a comprehensive data problem. Understanding data readiness is critical for success because accessible, high-quality data forms the backbone of accurate AI predictions.
The Hidden Costs of Preparing Data for AI at Scale
George Kurian points out that eighty-five percent of AI project time is consumed by preparing data before modeling begins. Data preparation involves cleaning, validating, and standardizing datasets to ensure accuracy and consistency across the organization. Without these steps, AI models may produce unreliable results that fail to meet business expectations.
Organizing data requires mapping information from multiple sources, aligning formats, and ensuring proper access controls are in place. Governance adds another layer, requiring policies that define who can use data and under what conditions. Enterprises often struggle to maintain these standards at scale, creating delays that extend project timelines significantly.
These tasks become even more complex when datasets are fragmented across departments, cloud platforms, and legacy systems. Engineers and data specialists must reconcile differences in structure, naming conventions, and missing values to create a coherent dataset. Even minor inconsistencies can cascade into major errors during AI training, forcing teams to repeat work and waste valuable time.
Kurian warns that organizations frequently underestimate the human effort and coordination required to make data usable. Preparing data is not a one-time activity; it demands ongoing maintenance and constant validation as datasets evolve. Failing to account for this hidden labor can derail AI projects before any meaningful insights are generated.
The cost of data preparation extends beyond time, impacting budgets, resource allocation, and project prioritization. Companies chasing model performance or GPU upgrades may overlook these foundational requirements, leaving AI initiatives vulnerable to delays. Ensuring comprehensive data readiness is essential to unlock AI’s full potential and prevent costly missteps.
Why AI Pilots Often Fail to Deliver Full Results
George Kurian highlights two major obstacles preventing AI pilots from scaling to full deployment. The first challenge is the inherent complexity of projects, particularly the extensive data preparation required upfront. The second challenge lies in organizational readiness and the ability of teams to adopt new workflows effectively.
Kurian emphasizes the importance of human change management when transitioning from pilot projects to enterprise-wide AI implementation. Engineers must learn to review code generated by AI instead of writing it entirely themselves. This shift in responsibilities requires training, clear communication, and cultural adaptation across technical teams. Without these measures, even successful pilot projects can stall and fail to provide business value.
Regions such as Korea demonstrate rapid adoption of new technologies but still face execution hurdles that slow AI integration. Kurian notes that public-private partnerships and fast adoption are strengths, but companies often underestimate the effort needed for companywide data alignment. Local firms may have advanced infrastructure, yet scaling AI demands consistent processes, governance, and interdepartmental cooperation. The speed of implementation alone does not guarantee successful AI deployment at scale.
Organizations must align technology stacks with clear business goals to overcome scaling obstacles. Fragmented internal and external datasets must be integrated to provide a full, actionable picture for AI models. Achieving this alignment requires executive sponsorship, cross-functional collaboration, and ongoing monitoring to ensure AI initiatives meet expected outcomes.
Kurian concludes that the most common misperception is treating AI as merely an infrastructure problem rather than addressing underlying data and organizational challenges. Companies that prioritize infrastructure over human readiness risk wasted investment and stalled projects. Success depends on a balanced approach that couples technological capability with comprehensive data strategy and workforce adaptation.
Three Priorities to Unlock Enterprise AI Value Quickly
George Kurian outlines three key priorities for organizations to maximize AI value efficiently across industries. The first priority is to experiment quickly, allowing companies to learn from failures and adjust strategies without large-scale risk. Firms that iterate rapidly often gain a competitive advantage by identifying effective AI applications before competitors.
The second priority is to align technology stacks with clear business objectives to ensure investments generate measurable outcomes. Organizations must evaluate how infrastructure, software, and data platforms support overall goals rather than treating AI as a standalone function. Clear alignment reduces wasted resources and increases the likelihood of successful deployment across sectors such as finance, healthcare, and manufacturing.
The third priority focuses on unifying fragmented internal and external datasets to provide a complete, actionable picture for AI models. Kurian emphasizes that AI models depend on high-quality, well-governed data for accurate predictions and reliable insights. Industries such as telecommunications, banking, and automotive frequently face challenges integrating siloed information, which NetApp helps address through enterprise data solutions.
Implementing these priorities allows companies to tackle both operational and data challenges simultaneously, strengthening the foundation for scalable AI initiatives. Combining rapid experimentation, strategic alignment, and data unification empowers teams to move beyond pilots and deliver tangible business results. Organizations that ignore any of these priorities risk underutilizing AI investments and limiting long-term growth potential.
NetApp’s role across multiple industries demonstrates how structured data platforms support these strategies in real-world contexts. From banks reviewing transaction patterns to manufacturers optimizing supply chains, AI success relies on coherent data strategies. By prioritizing experimentation, alignment, and data integration, enterprises can achieve value faster while reducing common risks associated with AI adoption.
When Data Outweighs Infrastructure in AI Investment Decisions
George Kurian’s central message underscores that AI success relies primarily on usable, accessible, and well-governed data. Companies often focus heavily on infrastructure upgrades while neglecting whether their data can support scalable AI models. Without a clear strategy for managing and unifying datasets, even the most advanced hardware will not deliver expected results.
Investing in AI infrastructure alone can create a false sense of progress while leaving critical data challenges unresolved. Organizations must treat data as a companywide asset, ensuring it is accurate, accessible, and compliant with governance policies. Firms that ignore this principle risk stalled AI projects, wasted resources, and missed opportunities to extract actionable insights.
Ultimately, viewing AI as a data problem rather than purely an infrastructure problem provides the foundation for long-term success. Aligning technology investments with comprehensive data strategies allows enterprises to fully realize the potential of AI applications. Kurian’s insight serves as a reminder that data readiness is the decisive factor in achieving sustainable, impactful AI outcomes.
