AI adoption is accelerating, but results are falling short, according to new DBTA research that will be released soon. Fewer than half of AI initiatives succeed, and most pilots never reach production.What's more, AI failures are expensive, with 36% costing more than $500K and 16% exceeding $1M.
The root cause is clear: data and the gaps in quality, access, trust, and lineage that continue to limit outcomes and prevent AI from scaling. This is not a new problem; it is an unsolved one.
At the same time, organizations are moving quickly into agentic AI, with 37% already in production, even as security, governance, data quality, and trust remain unresolved barriers.
Many are managing risk through copilots and human- in-the-loop workflows, but a clear gap remains: while most believe their data is AI-ready, most failures are tied to data, and only one in three can fully trace model outputs back to source data.
DBTA recently held a webinar, AI Readiness Reality Check: Why Most AI Initiatives Still Fail and How to Fix It, with John O’Brien, principal advisor and industry analyst at Radiant Advisors, Ryan Crochet, director of product marketing, and Bharath Vasudevan, vice president of product and GTM at Quest Software, for a data-driven discussion on what's really happening inside enterprise AI initiatives and why data readiness is the key to AI success or failure.
“When I started it was about getting your data house in order… but then AI came along, accelerated everything, and really raised the stakes,” Vasudevan said.
Although agentic AI has moved into production, converting pilots to production remains uneven, Crochet and Vasudevan noted. Most organizations rate themselves ready for AI, but outcomes tell a different story.
“This research study was really focused on AI-readiness and what it really means from an enterprise perspective,” O’Brien said. “We still have a healthy percent in the research and piloting phase. I think we’ve moved out of [2025] buzz and the big driver is ‘What can AI do for my company?’”
According to the soon to be released research survey, 74% of organizations rate themselves mostly or fully AI-Ready. Yet 52% report half or fewer of their AI initiatives succeeding and 77% of failures trace back to data. Even among organizations that rate themselves fully ready, only 51% achieve majority success. Confidence is high but the delivery is not, Crochet and Vasudevan said.
Only one in three can trace AI outputs back to source data, transformations, and business rules.
“AI is really built on that foundation of data readiness,” Crochet said.
For the full webinar, featuring a more in-depth discussion, Q&A, and more, you can view an archived version of the webinar here.