Ninety-five percent of artificial intelligence (AI) pilot programs fail to deliver measurable impact, according to research from MIT’s NANDA initiative.
This high failure rate indicates a significant gap between industry discourse on AI’s potential and actual implementation, with many organizations remaining in a pilot phase due to challenges with data management.
Organizations face difficulties managing the scale, complexity, and sensitivity of data required for AI development and deployment. Existing data resilience measures are often insufficient for an AI-driven environment.
Rick Vanover, Vice President of Product Strategy at Veeam Software, highlighted the central role of data in these challenges.
The global volume of data is projected to reach 181 zettabytes this year, tripling in five years, creating a data volume that exceeds many organizations’ current handling capabilities.
Gartner reports that 80% of enterprise data is unstructured, which historically limited its value extraction. AI technologies now enable organizations to derive value from this unstructured data.
The exponential growth of data, particularly with AI evolution, means companies struggle to categorize and manage their data. This issue exacerbates the problems encountered in AI pilot programs.
Despite aspirations for robust AI policies, “shadow IT” persists, with employees often experimenting with unauthorized AI tools due to organizational stagnation in pilot programs.
Effective data hygiene, including impact assessments, remains critical for AI implementation. Organizations must understand their data assets to identify integral information and ensure its resilience.
AI can assist in core data management tasks such as data classification, lineage improvement, and strengthening resilience measures. Prioritizing AI for data management can establish foundational control.
Organizations should initiate small, manageable AI projects that demonstrate value balancing innovation with control. This incremental approach builds confidence before attempting larger-scale transformations.
Continuous attention to the cost of creation, performance, and resiliency of AI models is necessary to build sustainable business processes around AI. Over-ambition without control risks operational resilience.
