The Marcus Aurelius Principle Your AI Center of Excellence Is Violating

Sixty percent of organizations have now established an AI center of excellence, and roughly 77% of their business units are quietly ignoring what it produces.

This is not a technology problem. It is not a change management problem. It is a philosophical one—and Marcus Aurelius diagnosed it nineteen centuries before the first enterprise AI framework was ever written.

In Meditations, Aurelius returns repeatedly to a distinction that Roman emperors and modern CoE leaders alike tend to collapse: the difference between appearing to do the work and doing the work. He wrote to himself, not for posterity. The Meditations were never meant to be published. They were a private accounting—a man holding his own conduct to the standard he claimed to believe in. He was not writing policy. He was practicing.

Your AI center of excellence, in all likelihood, is writing policy.

The Gap Between the Framework and the Floor

In conversations with enterprise teams, we observe a pattern so consistent it has become predictable: a CoE produces documentation that is technically rigorous, ethically considered, and operationally irrelevant. The business units that were supposed to follow it have already made their decisions by the time the guidance arrives—or they find the guidance so abstracted from their actual conditions that applying it requires more effort than proceeding without it.

The average gap between recognizing a problem and taking meaningful action is 14 months. Fourteen months is approximately how long it takes for a CoE to produce a governance framework, circulate it for comment, revise it, present it to leadership, and release it—by which point the problem it was designed to address has already metastasized or resolved itself by other means.

This is not incompetence. It is the natural consequence of building an institution whose primary output is guidance rather than capacity.

What Epictetus Would Tell Your CoE Director

Epictetus, whose teachings shaped Aurelius directly, organized everything around a single axis: the distinction between what is up to us and what is not up to us. He was not offering a comfort. He was offering a diagnostic.

Applied to an AI center of excellence: what is actually up to your CoE? It cannot control whether a marketing team uses an AI tool it discovered on its own. It cannot control whether a finance director waits for approved vendor lists or moves ahead with a pilot because quarter-end pressure is real and the approval queue is long. It cannot control the pace of the technology itself.

What is up to the CoE is whether the people it serves can do good work with less friction than they could without it. That is the whole question. Everything else—the framework documents, the governance tiers, the maturity models—is downstream of that answer.

When a CoE orients itself around producing artifacts rather than reducing friction, it has confused its means for its ends. Aurelius called this a failure of hêgemonikon—the governing faculty, the part of the mind responsible for direction and judgment. The institution has a body (staff, budget, mandate) but no governing faculty orienting that body toward what actually matters.

The 2 Questions That Expose the Architecture

There are two questions that will tell you, within a single conversation, whether your AI center of excellence is building expertise or building bureaucracy.

The first question: When a business unit last ignored your guidance, what did you learn from it?

An expertise-building CoE treats non-adoption as signal. It asks why the guidance failed to land, what conditions made it impractical, and what it needs to understand better about that unit's actual work. A bureaucracy-building CoE treats non-adoption as a compliance problem—something to be escalated, tracked, or reported to leadership as a risk.

The response to being ignored reveals everything about what the institution believes it is for.

The second question: Could a business unit do something faster, better, or with less risk because of what your CoE produced this quarter—and can you name the specific unit and the specific outcome?

Not a framework completed. Not a policy ratified. A unit. An outcome. A name.

If the answer requires more than thirty seconds, the CoE is measuring its own activity rather than its effect on others.

The Private Accounting

Aurelius did not write the Meditations to govern Rome. He wrote them to govern himself. The distinction matters enormously. Governance documents directed outward—at others, at institutions, at subordinates—produce compliance at best and resentment at worst. The discipline that actually moves things forward is the kind directed inward: holding yourself to account for whether your work produced anything real.

We see this reflected in a striking pattern: 67% of users who describe feeling stuck report that the stuckness predates their awareness of it by six months or more. The institution—like the individual—had already drifted before anyone named it. The frameworks kept being produced. The meetings kept being held. The activity continued while the actual purpose quietly receded.

The correction is not a new framework. It is a private accounting of the kind Aurelius modeled: direct, unsparing, and oriented entirely toward whether the work you claim to be doing is the work that is actually happening.

The Practice

If you lead, sit on, or report to an AI center of excellence, the philosophical move available to you right now is not strategic. It is investigative.

Find one business unit that has ignored your guidance in the past ninety days. Do not frame it as a compliance gap. Treat it as a data point about your own outputs. Ask them what they actually needed, what they did instead, and what would have made your guidance worth the effort of following.

Then write down what you learn—not for distribution, not for the next governance review, but for yourself. In the tradition Aurelius left us, that private accounting is not peripheral to the work. It is the work.

The CoE that produces guidance nobody follows has not failed at communication. It has failed at the prior question: what would make our work genuinely useful to the person doing the work. That question has no framework. It has only the discipline to keep asking it.