For a brief and optimistic period, the Center of Excellence felt like the future. 

Enterprises were discovering artificial intelligence with the earnestness of explorers, and the logic seemed impeccable. Intelligence was rare. Talent was scarce. Models were fragile and expensive. Of course it made sense to gather the brightest minds into a single place, give them the best infrastructure, and ask them to think on behalf of the rest of the organization. The Center of Excellence, or CoE, became the brain of the modern firm. Thought lived there. Standards lived there. Decisions lived there. 

What lived elsewhere was waiting. 

Today, that waiting has grown restless. 

Across industries and geographies, one hears a familiar complaint whispered in corridors and video calls: the CoE is slow. Or worse, it is brilliant but irrelevant. Requests disappear into queues. Job descriptions arrive already outdated. Models are elegant, yet curiously detached from the messiness of real work. Excellence accumulates, but execution feels stalled, as if intelligence has been preserved rather than applied. 

The irony is that this is not a failure of ambition. It is the delayed consequence of a structure designed for a different moment—one that even the most sophisticated business consulting firms are now questioning. 

Because while organizations were perfecting their Centers of Excellence, intelligence itself was changing shape. 

The Moment Intelligence Left the Tower 

Artificial intelligence no longer behaves like a rare resource that must be protected. It behaves more like electricity. Or perhaps more accurately, like eyesight. 

The most consequential systems emerging today do not simply analyze data or generate text. They observe. They watch workflows unfold. They read documents the way people do, scanning for meaning rather than structure. They look at screens, at images, at video feeds, and then decide what to do next. These “agents,” as they are now called, do not wait for permission in the same way earlier systems did. They are built to act. 

This shift has unsettled an unspoken assumption that governed the first wave of enterprise AI: that intelligence must be centralized in order to be safe. 

In an agentic world, that assumption quietly collapses. 

The most valuable intelligence is no longer the most sophisticated. It is the most contextual. The people closest to customers, to operations, and to friction-filled processes now possess something the smartest centralized team cannot replicate: proximity to reality. When intelligence can be embedded directly into work, the question is no longer whether the center is capable, but whether it is necessary as a gatekeeper. This is a question that top IT strategy consulting firms and growth strategy consulting firms are helping their clients navigate with increasing urgency. 

Many organizations sense this tension but struggle to name it. They feel it when business teams build unofficial automations on the side. When “shadow AI” proliferates not because people are reckless, but because they are tired of waiting. Control tightens in response. Governance committees multiply. And yet, risk increases rather than diminishes, because the center has mistaken authority for safety. 

The Strange Comfort of Control 

Centralization has always offered psychological comfort. Leaders know where decisions live. Accountability feels clearer. Risk appears manageable. 

But comfort is not the same as resilience. 

In complex systems, excessive centralization creates fragility. When everything must pass through a single point, that point becomes the slowest, most overloaded, and most politically charged part of the system. The CoE, once envisioned as a brain, begins to resemble a bottlenecked nervous system, overwhelmed by signals it was never designed to process at scale. 

What emerges instead is a peculiar organizational theatre. The rest of the enterprise is sent to “learn AI.” Workshops are run. Certifications are issued. Fluency is encouraged. And yet, the authority to act remains concentrated elsewhere. People become students of intelligence rather than its practitioners. Learning becomes an end state. Agency remains theoretical. 

This is how schooling replaces capability. 

And it is why the next phase of enterprise AI will not be defined by better models, but by a redistribution of power—a shift that will fundamentally reshape tech strategy consulting and how organizations approach revenue growth management. 

The Quiet Resignation No One Talks About 

Every major organizational shift requires a resignation, though rarely a formal one. 

In this case, it is the resignation of centralized excellence as the sole owner of intelligence. Not because it lacks value, but because its role must change. The center must stop being the place where intelligence is made and start being the place where intelligence is made safe to use. 

This is a subtle but radical redefinition. 

Instead of deciding what gets built, the center designs the conditions under which many things can be built. Instead of approving every decision, it embeds guardrails so that decisions can be made locally without fear. Governance, once synonymous with delay, becomes invisible infrastructure. Like good medicine, it works best when it is barely noticed. 

Some organizations have begun to describe this shift with a metaphor borrowed from biology: the flower. 

In a flower, the stem does not dominate the petals. It supports them. It provides structure and nourishment, but growth happens outward, in many directions, responding to light and context. The petals are where reproduction occurs. Pollen travels. Capability spreads. 

It is a more honest description of how intelligence wants to behave today. 

When Governance Stops Causing Headaches 

One of the most persistent myths in enterprise technology is that speed and safety are opposites. 

They are not. 

They are simply designed at different layers. 

When governance relies on meetings, reviews, and manual approvals, it inevitably slows things down. When it is built into systems—through automatic redaction, decision logging, policy enforcement, and kill switches—it becomes something else entirely. It becomes relief. Ibuprofen for the organizational headache. 

Teams move faster not because they are less cautious, but because they no longer have to carry the cognitive burden of risk themselves. Safety becomes ambient. Confidence replaces fear. And the incentive to work in the shadows quietly disappears. 

This is the moment where centralized control gives way to distributed trust—not blind trust in people, but trust in systems designed to absorb error without collapsing. 

The End of “AI Training” as We Know It 

Perhaps the most underestimated consequence of this shift lies in how organizations think about talent. 

For years, the response to AI has been education. Literacy programs. Upskilling initiatives. Workshops that promise fluency. These efforts are not misguided, but they are incomplete. Learning without authority creates spectators, not builders. 

In an agentic enterprise, capability is not proven in classrooms. It is proven in production. An agent that runs. A workflow that reduces friction. A decision that happens faster and better than before. 

This changes the meaning of expertise. Intelligence stops being a specialist possession and becomes a shared layer of competence. Much like financial judgment or operational discipline, it becomes something leaders are expected to exercise, not outsource. 

Why the Edges Will Win 

There is a reason emerging enterprises, particularly those accustomed to federated delivery and contextual execution, appear more comfortable with this future. They were never fully centralized to begin with. They learned to operate with trust, distance, and imperfect information. 

For them, the agentic era is not a disruption. It is an acceleration. 

The real divide will not be between companies that adopt AI and those that do not. It will be between organizations that cling to intelligence as a controlled asset and those that allow it to travel. 

Because intelligence, it turns out, does not want to live in towers. 

It wants to sit beside the work. To watch. To act. To move on. 

And the enterprises that learn to let it do so will look less orderly at first. They will also learn faster, adapt better, and quietly outpace the rest. 

A flower does not fear the loss of its pollen. 

It understands that this is how life spreads.