Why AI Transformations Fail
Published on: 01/06/2026
Roughly 80% of enterprise AI initiatives never reach production. From what we've seen across $1B+ in transformations, here's why — and what it takes to avoid each failure mode.
BusinessAI
AI & Business
Field notes from inside enterprise AI.
Strategy, build, and execution. Written by the EKP team and the operators we work with.
Industry Reading
Additional insights from the industry.
What we're reading and how we're thinking about it. Research and commentary from outside our team, paired with our take.
RAND Corporation
The root causes of failure for AI projects and how they can succeed
While the MIT findings focus on generative AI pilots specifically, a 2024 RAND Corporation study took a wider lens, examining why AI projects across the board fail to reach production. Drawing on interviews with data scientists and engineers, RAND found that the majority of AI initiatives never ship, at a rate well above that of conventional IT projects. The notable part isn't the number; it's the cause. The failures RAND identifies are rarely about model performance. They trace back to misunderstood problems, missing or mishandled data, infrastructure that wasn't ready, and a focus on the technology rather than the business problem it was meant to solve, the same structural patterns we see in the field, and have solutions to.
Read the full study
Yahoo Finance
Greg Abel reveals Berkshire's 'Narrow AI' direction
Greg Abel, Warren Buffett's successor atop the most disciplined capital allocator in the world, has framed Berkshire's approach to AI in characteristically pragmatic terms: not chasing the technology for its own sake, but applying it narrowly, where it solves a real problem and creates durable value. It's the opposite of the hype cycle. That's the same philosophy Emerald Key is built on: find the places where AI moves the business, build those well, and ignore the rest. When the world's most patient capital treats AI this way, it isn't caution, it's the discipline that tends to win.
Read the full article
McKinsey & Company
Superagency in the workplace: Empowering people to unlock AI's full potential
This latest report by McKinsey & Co. emphasizes the need for leaders to act boldly and responsibly to harness AI's full potential and drive transformative change in the workplace. Starting now and making sure you correctly navigate the short-term complexities to get off the ground will be important to ensure you are part of the long-term potential and not left behind.
Read the full article
Fortune
MIT report: 95% of generative AI pilots at companies are failing
Enterprises that are hastily grabbing off-the-shelf tools or surface-level proofs-of-concept are failing at a high rate. To capture the true value of AI, it must be approached thoughtfully and be a transformative step where AI is integrated with your true business processes. Otherwise, results can't be trusted or replicated, adoption never takes hold and impact isn't sustained. The full impact of AI is realized through agents, applications and tools with specific skills, calibration and training embedded to give a competitive edge.
Read the full article