AI isn’t failing; your enterprise systems are (www.techradar.com)

A recent analysis highlights a startling statistic: approximately 80% of AI projects fail to deliver tangible business value, a sentiment backed by a 2025 MIT study that reported 95% of generative AI pilots yielding no measurable return. The common instinct is to attribute these failures to AI models themselves; however, experts argue that the real issue lies within the complex, outdated enterprise systems they operate in. Many organizations rely on tangled infrastructures, consisting of mismatched data structures and inconsistent information that invariably hinder AI’s success. For AI to realize its full potential, organizations must prioritize the establishment of robust data infrastructure and seamless system integrations. This entails cleaning up messy data, standardizing inputs, and ensuring that AI outputs are effectively integrated into daily workflows. Furthermore, deploying AI requires a multidisciplinary approach that includes not just data scientists and software engineers, but also a strategic focus on organizational structure and team composition. Companies that excel in harnessing AI typically do the foundational work first—creating organized, connected systems that support automation and deliver real business value—rather than simply adopting powerful models without addressing the underlying issues.