Why AI Transformations Are Failing—And What CEOs Can Do

Reports keep piling up with the same story: Enterprise AI adoption is failing to deliver on its promise. Billions are being spent, yet the results look almost identical to past transformation failures.

Why? Because CEOs are making the same mistake again: Focusing too much on the tech—AI in this case—rather than the people, teams and systems that must adapt to use it.

Your AI program should be treated like a high-value product: offered to teams, not mandated. It deserves the same resource support and operational rigor as any of your top-performing products. That means clear ownership, strategic leadership and, above all, program team accountability for measurable results. In this model, AI is the product, your organization is the marketplace and your teams are the customers.

That’s how we approached IBM’s global transformation in the 2010s. We helped thousands of interdisciplinary teams become more entrepreneurial, agile and customer-focused. Tools and processes were just part of the story. The real shift was cultural—changing how people worked together. Those changes stuck, and they remain core to how IBM operates today.

From that experience, I see three lessons for CEOs who want to capture AI’s promise and make it stick.

1. Your Change Product

• Brand it with values. People don’t rally around tools; they rally around meaning. At IBM, we didn’t call our program “design thinking”—and you shouldn’t call yours “AI.” Too much baggage, too narrow a remit. Instead, we branded it Hallmark, a neutral vessel infused with values about collaboration and client impact. That framing made the effort bigger than any single tool.
  
  

• Build the right leadership team. A product needs its own leadership team—cross-functional, empowered and accountable for adoption. Not just evangelists, but leaders with operational skills like communications, sales and finance, connected directly to scaled functions such as the CFO, CIO and CHRO.
  
  
• Treat teams as your users. Think of your organization as the marketplace, and intact teams—not individuals—as the customers. Individuals influence, but teams deliver outcomes.

2. Adoption, Not Enablement

• Make training team-specific. Generic workshops rarely drive adoption. Each team needs a tailored entry point—and the freedom to walk away if the “product” isn’t better than their status quo.
  
  
• Measure outcomes, not inputs. Counting class attendance or the percentage of code or emails generated doesn’t reveal anything of significance. The real metric is whether teams adopting the changes are delivering better outcomes. At IBM, we had dedicated program team members embedded with each Hallmark team for up to six months. Of course, they supported the teams, but their real purpose was to understand what was and wasn’t working so that the program team could make changes to the product.
  
  
• Put a price on it. Free signals low value. At IBM, teams had to budget for inclusion in the Hallmark program, which included paying for tools and people. If a team’s leader wasn’t willing to invest, that told us the product wasn’t compelling enough. Charging created accountability: it forced the program team to deliver real value, rather than blaming business units for “not getting it.”

Scaling

• Start small, then scale. Begin with a handful of teams to expose cultural friction points and system barriers. Fix those before rolling out broadly.
  
  
• Find the magic people. Every organization has them—team members who naturally bring others along. Baseball teams call these people “glue guys.” We identified, supported and often redeployed them onto struggling or new teams to spread belief and momentum.
  
  
• Equip and protect middle managers. Our best teams thrived because managers shielded them from counterproductive directives—we called them “sh*t umbrellas.” From them, we learned to train managers differently: not on the product itself, but on how to manage teams using the product. Once we updated incentives and badged these managers as “Advocates,” adoption accelerated dramatically.

Conclusion

AI transformations aren’t failing because the technology is flawed. They’re failing because leaders are thinking more about tech than about the cultural assimilation.

If you’re serious about making your AI-based transformation stick:

• Offer your AI program as a product to a few teams willing to make it succeed—ones that will be transparent but not judgmental about what works and what doesn’t.
• Learn from those early teams where systems and processes thwart adoption, and fix them.
• Then scale, tracking key people—your magic people and advocates—who can accelerate the shift.

That’s how IBM’s cultural transformation became something 400,000 people chose to adopt. And it’s how your organization can create an irresistible appetite for AI-based transformation that sticks.