Article 3 in a four-part series on AI governance for operational leaders. Read Article 1 and Article 2 at rachelkimber.substack.com
There is a question I put to every operational leader I work with.
Not at the beginning. You have to earn this question. It only lands after someone has done the work of looking honestly at their organization — the gaps, the unowned workflows, the policies drafted in good faith that practice has already lapped — and started building something real.
The question is this:
If you left tomorrow, would your organization know what to do?
Not: would they be okay eventually. Not: could they figure it out. Would they know — on day one, in the specific situation, with the specific staff member who has a specific AI tool open and a specific deadline looming — what the right decision is.
Most leaders pause.
That pause is the governance work.
Governance that only holds when you are in the room to explain it is not governance. It is individual judgment.
The Bicycle
The third session of the AI Governance & Operational Leadership Cohort is called The Bicycle.
By Session 3, participants have already done two hard things. In Session 1 — The Mirror — they looked honestly at what AI is revealing about their organizations. The places where accountability is fuzzy. The decisions that have always depended on one person’s institutional memory. The values on the website that don’t show up in Tuesday afternoon choices.
In Session 2 — The Intern — they built something. A practice guideline for one real AI use case. Named ownership. A review step that is real, not performative. The supervision structure that transforms AI from an unsupervised capability into a managed one.
Session 3 asks whether what they built can move without them.
A bicycle does not stay balanced because someone holds it upright. It stays balanced because of how it was built and because the rider has internalized the skill. Forward motion creates stability. The system works when you let go.
That is what practice guidelines are building toward. Not a document you have to explain. A structure that holds because of how it was designed — not because of who is watching it.
What Actually Breaks
I have been doing AI governance work with foundations and nonprofits long enough to know where implementation fails. It is almost never where people expect.
It is not usually a rogue staff member or a catastrophic data breach. It is the quiet, cumulative gap between a policy that exists and a practice that holds.
Three patterns show up consistently.
Analysis paralysis. Waiting until the complete policy, the enterprise risk framework, or the approved tool list is finished before governing anything. Meanwhile, staff are experimenting. Tools are being adopted. Decisions are being made without guidance, not because leaders don’t care — because they are waiting to get it right before they start.
Accountability fog. An AI committee that advises but does not own. A practice guideline that describes a review process without specifying who carries it out. A governance document that lands in the shared drive and stays there because no single person is accountable for whether it is used.
COO bottleneck. Governance that lives in one person’s head. Every question comes back to the same leader. Every exception requires their judgment. The organization is not governing AI — it is relying on one person’s presence to substitute for governance. When that person is unavailable, in a board meeting, traveling, or simply not yet copied on the email, the structure doesn’t hold.
Every leader I work with recognizes at least one of these in their own organization. Naming them is not pessimism. It is the diagnosis that makes the design possible.
The Ownership Gap
The most recurring finding in the cohort — more than data risk, more than tool selection, more than any specific AI capability — is ownership.
Who decides. Who reviews. Who is accountable when something goes wrong.
In Session 3, participants pressure-test each other’s practice guidelines. The instruction is direct: be critical. Find the holes before a staff member does.
What comes back consistently: roles described without names attached to them. Review processes that assume someone will carry them out without specifying who. Human decision fields with no single person in the box — just a cluster of overlapping responsibilities and no one with full authority.
The COO was the de facto owner of everything.
Which is not ownership. It is exposure.
The fix is not complicated. It requires naming someone — a specific person — who is accountable for each practice guideline. Who owns it. Who updates it. Who a staff member calls when the situation doesn’t fit the guidance.
Once you name owners, something shifts. The accountability stops being entirely yours. And governance starts becoming organizational rather than individual.
Values Are Not Decorative
There is a section in the practice guideline template that most leaders initially treat as a formality.
The values section.
Session 3 is where that changes.
One participant in the cohort had just completed a two-year organizational rebrand. Four values. Staff recite them from memory at every all-hands. She described realizing — mid-session, for the first time — that she could run AI governance decisions directly through those values. Not as a framing device. As actual decision criteria.
Act with kindness: clarity is a form of kindness. A practice guideline that tells staff exactly what to do is not bureaucracy. It is care.
Another participant grounded her transcription guideline in co-creation and action — two of her organization’s core commitments. Her argument: asking someone to take notes while also asking them to collaborate fully in a meeting is a contradiction. AI transcription resolves that contradiction in a way that is consistent with the organizational commitment to full, equitable participation.
That is not a convenience argument. It is a values argument.
Values arguments land differently with staff. They land differently with boards. They are the reason a practice guideline holds when you are not in the room — because staff are not following a rule. They are acting from a shared understanding of what the organization actually believes.
That is how the bicycle stays balanced.
The Question This Series Has Been Building To
Article 1 established that governance is never finished. The moment you articulate a guideline, AI is already shifting the workflow underneath it. The baseline is not the ceiling. It is the floor.
Article 2 established that AI governance is a management problem, not a technology problem. The intern isn’t the problem. The absence of a manager is. You already know how to supervise. You just haven’t applied that knowledge here.
Article 3 — this one — is asking whether what you’ve built can travel. Whether it holds at 7am before you log on. Whether a new hire could follow it without your explanation. Whether it survives a leadership transition.
These are not abstract questions. They are the test of whether governance is real.
And they point toward something larger than any one practice guideline.
The organizations navigating AI responsibly right now are not the ones with the most comprehensive policies. They are the ones building institutions capable of making good decisions when the policy runs out. When the situation is new. When the tool changes. When the staff member has to make a call and the only thing guiding her is what she has internalized about what the organization actually values.
The strongest organizations are not creating perfect AI policies. They are building institutions capable of making good decisions when the policy runs out.
This Is the Room
The AI Governance & Operational Leadership Cohort is for COOs, executive directors, and senior operational leaders who are somewhere between experimenting and governing.
Not waiting for the perfect moment. Not looking for a training or a template. Ready to do the actual work — for one real decision, in their organization, with peer leaders who are navigating the same terrain from different angles.
Six leaders. Four sessions. One real decision carried the full arc from inventory to governance standard to 90-day commitment. Philanthropy seats underwrite nonprofit participation. Power-balanced learning is structural, not incidental.
What participants leave with is concrete: a mapped governance snapshot, a pressure-tested practice guideline, named ownership, a defined disclosure posture, and a 90-day implementation pathway specific to their organization’s culture and capacity.
What they also leave with is harder to put in a list. The clarity that comes from doing the work in community. The relief of realizing that governance is not a technology problem — it is a management problem, and you already know how to manage. The confidence of having said a specific commitment out loud to peers who will remember you said it.
The next cohort is forming now.
If your organization is somewhere between a policy that exists on paper and a practice that holds — this is the room.
Applications for the next cohort are now open.
If this is the work your institution needs, let’s talk.
Apply HERE
Let’s chat about all other things HERE