Loading content...
Jonathan Hansing
The Real Cost of AI Transformation
Most AI initiatives fail not because of technology, but because organizations underestimate the true scope of transformation required across people, processes, and technology.
Most AI initiatives fail not because of technology, but because organizations underestimate the true scope of transformation required across people, processes, and technology.
Last month, a CEO told me his board wanted him to "sprinkle AI on" the business. His words, not mine. When I asked what problem they were trying to solve, he paused. "I think they just don't want to get left behind."
This is where most AI conversations start. And it's exactly why most AI initiatives stall.
A RAND Corporation study found that more than 80 percent of AI projects fail. That's twice the failure rate of IT projects that don't involve AI. The technology isn't the problem. The organization is.
The enthusiasm is real. Every executive I talk to wants AI in their business. But wanting AI and wanting what AI actually requires are two different things. The gap between "yes, let's do AI" and "yes, let's transform how our people work, how our processes flow, and what our technology stack looks like" is where ambition goes to die.
Here's what usually happens. A leadership team gets excited about AI's potential. They greenlight a pilot. The pilot works. Then someone asks what it would take to scale it across the organization, and the room goes quiet.
Scaling AI isn't a technology problem. It's a transformation problem. And transformation touches everything.
People need to think differently. Processes need to be redesigned, not just automated. Technology that seemed "good enough" suddenly isn't. The bill comes due not in software licenses but in organizational change, and that's when enthusiasm meets hesitation.
This hesitation isn't weakness. It's the correct response to realizing the actual scope of the work. The problem is when hesitation becomes paralysis.
Scott Anthony, Clark Gilbert, and Mark Johnson offer a useful frame for thinking about this challenge in Dual Transformation. Their insight: companies facing disruption need to pursue two transformations simultaneously.
To succeed at the difficult task of dual transformation, leaders must have the courage to choose well before the platform burns; the clarity to focus on a few high-potential opportunities; the curiosity to explore, even in the face of probable failure; and the conviction to persevere through crises of conflict, identity, and commitment.
— Scott Anthony, Clark Gilbert, and Mark Johnson, Dual Transformation
Transformation A repositions today's business for resilience. You're not abandoning what works. You're making the core stronger, more efficient, more defensible. Think Adobe shifting from packaged software to subscription services. Same customers, same problems, fundamentally different delivery.
Transformation B creates tomorrow's growth engine. This is the new business model, the new value proposition, the thing that doesn't exist yet but will define your future.
The key word is simultaneously. Not sequentially. Not "fix the core first, then innovate." Both at once, which is exactly as hard as it sounds.
For AI transformation, this framework clarifies the work:
Transformation A means embedding AI into your existing operations. Automating what can be automated. Augmenting what should be augmented. Making your current business faster, smarter, more responsive.
Transformation B means building new capabilities that AI makes possible. New products. New services. New ways of creating value that weren't feasible before.
Most companies attempt one or the other. The ones that pull ahead do both.
The talent question isn't "how do we hire AI experts?" It's "how do we make everyone AI-capable?"
This doesn't mean teaching your sales team to build neural networks. It means building a workforce that understands what AI can do, knows how to work alongside it, and can spot opportunities to apply it. Generalist systems-thinkers who can connect dots across domains will matter more than narrow specialists who can't see beyond their function.
The cultural shift is harder than the technical one. You need an organization where experimentation is normal, where failure is data rather than career risk, where people feel safe trying new approaches. Yet most organizations underinvest here. Deloitte's State of AI in the Enterprise research found that only 37% of organizations make significant investments in change management, incentives, or training to help people integrate new technology into their work.
The payoff for getting this right is substantial:
Organizations that invest in change management are 1.6 times as likely to report that AI initiatives exceed expectations.
— Deloitte, State of AI in the Enterprise
Leaders set this tone. If the C-suite treats AI as something the IT department handles, everyone else will too. But if executives are visibly learning, visibly experimenting, visibly struggling with the same tools they're asking others to adopt, that's a different signal entirely.
Incentives follow attention. What gets measured, rewarded, and celebrated shapes behavior faster than any training program.
McKinsey's research on AI change management reinforces this point:
Companies involving at least 7 percent of employees in transformation initiatives double their chances of delivering positive excess total shareholder returns, with the highest performers involving 21 to 30 percent of employees.
— McKinsey, "Reconfiguring Work: Change Management in the Age of Gen AI"
This isn't a top-down mandate. It's a broad mobilization.
Here's a trap I see constantly: wrapping AI around a bad process.
You've got a manual workflow that takes too long, frustrates everyone, and produces inconsistent results. AI can automate it. Suddenly the bad process runs faster. Congratulations. You've made dysfunction more efficient.
AI doesn't fix broken processes. It accelerates them. If your process has unnecessary steps, poor handoffs, or unclear ownership, automation will execute those flaws faster and wider than any human team could.
Lisanne Bainbridge identified this paradox in her classic 1983 paper on automation, and it applies directly to AI:
One might state these problems as a paradox, that by automating the process the human operator is given a task which is only possible for someone who is in on-line control.
— Lisanne Bainbridge, "Ironies of Automation" (1983)
In other words: the people best equipped to catch AI mistakes are the ones whose skills will atrophy fastest once AI takes over the work. You can only monitor effectively if you deeply understand the work being done. But if AI is doing the work, you're no longer building that understanding.
This is where design thinking matters. Before you automate, you need to ask: Should this process exist at all? What's the actual outcome we need? What would we build if we were starting from scratch?
The processes most ripe for AI transformation are often the ones no one has questioned in years. They're "just how we do things." But tradition is not a strategy, and "we've always done it this way" is not a reason to keep doing it.
Your current tech stack was probably built for a different era. Systems that have worked fine for years suddenly become bottlenecks when you try to layer AI on top of them.
The C-suite knows this is a problem. Research published in MIT Sloan Management Review found that 72% of executives say technical debt greatly limits their ability to migrate to new technologies, and 70% say it severely limits their IT function's ability to innovate. The kicker? 67% want to replace all their legacy systems, but 70% also want to keep them as long as possible. That's not a strategy. That's paralysis.
The question isn't whether you need to modernize. It's how.
Full stack replacement is rarely realistic. The cost, the risk, the organizational disruption would sink most AI initiatives before they start. But you can't ignore the problem either.
The answer, for most organizations, is rolling migration. You don't replace everything at once. You identify the systems that most constrain your AI ambitions and prioritize those. You build bridges between old and new. You accept that you'll be running hybrid environments for a while.
One non-negotiable: if your systems don't have APIs or easily accessible data, they need to go. AI runs on data. If your data is trapped in formats that can't be extracted, connected, or analyzed, you don't have an AI problem. You have a data problem that's blocking everything downstream.
The worst thing a leader can do is launch AI initiatives without honest assessment of organizational readiness.
Capability matters: Do you have the skills, the systems, the data infrastructure to execute?
Appetite matters too: Is there genuine willingness to change how people work, to redesign processes, to retire legacy systems?
Capability without appetite produces pilots that never scale. Appetite without capability produces chaos. You need both, and you need them at levels that match your ambitions.
This assessment isn't a one-time exercise. As you learn what AI transformation actually requires, your understanding of both capability and appetite will evolve. The goal isn't perfect knowledge before you start. It's honest knowledge about where you are so you can plan intelligently.
AI transformation is not a technology project. It's not even a digital transformation project. It's organizational transformation that happens to be triggered by technology.
The RAND researchers who studied why AI projects fail put it bluntly: the most common causes aren't technical. They're misaligned goals, inadequate data infrastructure, and organizations chasing shiny technology instead of solving real problems. The fix isn't better algorithms. It's better alignment between what leaders say they want and what the organization is actually prepared to do.
But here's what I can't tell you: how long this takes, or whether the gains will justify the cost for your specific organization. The companies that succeed will pursue both transformations at once, invest in people as much as platforms, redesign processes before automating them, and make hard choices about technology debt. But "succeed" is doing a lot of work in that sentence. Transformation is not a destination. It's a permanent condition.
"Sprinkle AI on it" isn't a strategy. Neither is "transform everything." The real work is finding what's true for your organization, with your people, given your constraints. That's harder than any framework can capture. And it's exactly the work that can't be automated.
The honest assessment of where you stand, not where you wish you were, is the first step. We built a tool to help with that: a free assessment that evaluates your organization across both Capability and Appetite. Fifty questions, fifteen minutes, and you'll know whether you're a Believer, a Skeptic, a Pragmatist, or a Leader. Take the AI Maturity Assessment →