AI is the new cloud — and we’re repeating the same mistakes

A few years ago, I sat through countless meetings where leaders debated whether their organizations were ready for cloud computing. Security teams worried about risk. Executives worried about cost. Engineers worried about migration complexity. Everyone was focused on the technology.

Today, I hear many of the same conversations about AI.

The technology has changed. The underlying problem has not.

Over the course of my career, I’ve helped lead modernization efforts across military, government and enterprise environments, including cloud migrations, enterprise application transformations and operational technology programs. Those experiences taught me that technology is rarely the hardest part of transformation. One lesson has remained remarkably consistent: technology is rarely the hardest part of transformation.

People are.

Processes are.

Operations are.

That’s why I believe many organizations are approaching AI from the wrong direction. They’re treating it as a technology initiative when it’s actually an operational transformation.

Much like the early cloud era, organizations are racing to evaluate models, pilots and vendors. Yet many have spent surprisingly little time examining the workflows, decision paths and organizational structures those technologies will ultimately support.

As CIOs learned during previous platform shifts, technology alone doesn’t create transformation. Operational readiness does. In many ways, today’s AI discussions remind me of CIO’s article, “ Your cloud provider is a single point of failure_.”_ The technology decision may be important, but the operational dependency it creates is often far more significant.

Most organizations don’t have an AI problem. They have an operational readiness problem

When AI initiatives struggle, the first instinct is often to blame the model.

Maybe the outputs aren’t accurate enough. Maybe the prompts need improvement. Maybe the organization selected the wrong vendor.

In my experience, those issues are rarely the primary reason initiatives stall.

The real obstacles tend to be much less exciting:

• Unclear ownership
• Poorly defined business processes
• Inconsistent data sources
• Lack of trust in outputs
• Undefined approval authorities
• Competing organizational priorities

During one large-scale modernization effort, I watched teams spend months debating infrastructure choices while spending almost no time redesigning the operational processes the infrastructure would support. The technology worked exactly as intended. Adoption became the bottleneck.

What has struck me recently is that the conversation has become remarkably similar across organizations. Regardless of industry, leaders are asking many of the same questions: Where does AI fit? Who owns it? How do we introduce it without disrupting critical operations?.

In other words, many organizations want AI. Far fewer are operationally ready for it. One reason this readiness gap concerns me is that many organizations are approaching AI as though it were a standalone capability. In reality, AI sits atop everything that already exists. It depends on data, processes, governance structures, security controls and organizational trust.

I’ve spent much of my career working in environments where technology changes cannot simply be pushed into production and evaluated later. Systems support real business operations, regulatory requirements and mission outcomes. Even relatively small changes often require coordination across technical teams, security stakeholders, business owners and leadership.

AI introduces an additional layer of complexity by influencing how decisions are made. That means organizations need to think beyond implementation and consider adoption, accountability and oversight from the beginning.

For example, if an AI capability recommends a course of action, who is responsible when that recommendation is accepted? What happens when the recommendation conflicts with established procedures? How should organizations document decisions that were influenced by AI? These questions often emerge after deployment, but they should be addressed before deployment.

The organizations making the most progress with AI are not necessarily the ones moving the fastest. In many cases, they are the ones spending time upfront to understand how AI will fit into existing operational models. They recognize that success depends not only on what the technology can do, but also on whether the organization is prepared to use it responsibly and consistently.

Organizations spend enormous energy evaluating capabilities while investing far less effort in understanding how decisions are actually made inside their organizations. The latest McKinsey State of AI research continues to show that while AI experimentation is widespread, many organizations still struggle to generate sustained business value from those investments.

Before deploying AI, leaders should be able to answer some basic questions:

Who owns the process being automated?

How are decisions currently made?

Where does human oversight belong?

What happens when the AI recommendation is wrong?

If those questions cannot be answered clearly, AI will not solve the underlying problem.

In fact, AI may simply help the organization execute a broken process faster.

Governance is necessary, but it isn’t enough

As someone who has spent significant time working in regulated environments, I understand why governance has become a major focus of AI discussions.

Frameworks such as the NIST AI Risk Management Framework provide valuable guidance for managing risk and establishing accountability. But governance alone does not create operational readiness.

This is where many organizations get stuck.

Governance is frequently treated as a separate activity performed by legal teams, compliance offices or security organizations. Operational teams often view governance as something that happens after the real work is complete.

That approach rarely succeeds.

In successful transformations, governance becomes part of the operational process itself.

If an AI system is generating recommendations that influence decisions, organizations should already know:

• Who reviews those recommendations
• What authority that person has
• How decisions are documented
• How exceptions are handled
• How errors are identified and corrected

Those aren’t primarily technical questions.

They’re operational questions.

In many ways, AI governance resembles change management more than software engineering. Success depends on creating clear roles, responsibilities and decision paths that people understand and trust.

The organizations that integrate governance directly into operational workflows will move much faster than those that treat governance as a separate layer of oversight.

That challenge becomes even more important as organizations begin experimenting with increasingly autonomous systems. As CIO explored in “ The AI revolution: Getting culture right for AI success_,”_ technology adoption ultimately succeeds or fails based on culture, trust and organizational alignment rather than technical capability alone.

The next generation of AI leaders will be operators

Much of today’s AI conversation focuses on technical expertise.

Technical expertise matters. Organizations need talented engineers, data scientists and architects.

But I believe the next wave of AI leadership will come from operators.

The leaders who create lasting value from AI will be the people who understand how work actually gets done inside complex organizations. They will understand organizational behavior, risk management, accountability and decision-making. They will know where automation creates value and where human judgment remains essential.

Most importantly, they will recognize that technology alone rarely transforms anything.

I’ve seen organizations struggle with ERP deployments, cloud migrations and enterprise modernization efforts. In almost every case, the limiting factor wasn’t the technology itself. It was the organization’s ability to adapt its processes, responsibilities and ways of working.

AI will be no different. One lesson I’ve learned from previous technology transformations is that organizational patience often runs out before organizational readiness arrives. Leaders feel pressure to demonstrate progress, employees worry about disruption and technology teams are asked to move faster than the surrounding processes can support.

AI is creating similar pressures today. The excitement is understandable. The opportunities are real. But organizations that focus exclusively on deployment metrics may find themselves struggling with adoption, trust and accountability later.

In my experience, sustainable transformation happens when leaders are willing to invest as much effort in operational preparation as they do in technology selection.

The organizations that succeed won’t necessarily have the most advanced models.

They’ll have the clearest ownership, the strongest operational discipline and the highest levels of trust.

That’s why I increasingly view AI not as a technology challenge, but as an operational one.

We already know how technology transformations fail. Many of us have watched it happen before.

The organizations that remember those lessons will have a significant advantage as AI moves from experimentation into everyday operations.

This article is published as part of the Foundry Expert Contributor Network. Want to join?