COHORT REFLECTION | FULL CIRCLE IMPACT SOLUTIONS | ARTICLE 2 OF 4
Article 2 in a four-part series on AI governance for operational leaders. Start with Article 1 — The Guideline Moved the Moment We Wrote It
Session 2 of the AI Governance & Operational Leadership Cohort was called The Intern.
I named it that deliberately. Because by the time leaders arrive at Session 2, they have already done the mirror work. They have looked honestly at what AI is revealing about their organizations — the gaps between aspiration and practice, the workflows nobody owns, the policies that exist on paper while staff experiments happen in real time, unchecked and ungoverned.
Session 2 is where they have to actually do something with what they saw.
The room came in carrying it. Overwhelmed. Scattered. Tired. One participant had just finished extensive travel. Another came straight from a week with her board. The energy was not resistant. It was genuinely taxed.
And underneath the tiredness was something more specific: a growing awareness that the governance question was not going to be resolved by reading more articles or attending more webinars.
They had to build something. For real. For one specific workflow. Right now.
That is what The Intern is for.
“You do not hand a new intern your most sensitive work and walk away. You tell them specifically what they are allowed to draft, what gets reviewed before it goes anywhere, and what they do not touch. Not because the intern is untrustworthy. Because exquisite clarity is how you build trust over time.”
AI works the same way.
The governance question is not whether to use it. It is what it can do, what gets reviewed, and what only a human decides.
The Most Capable Intern You’ve Ever Hired
Here is the thing about a brilliant intern. They are genuinely impressive. Fast. Eager. Tireless. They will take on whatever you hand them without complaint, without fatigue, without asking for a long weekend.
The AI-Intern is also — and this is the critical part — brilliant at pattern recognition and completely without common sense.
They will write you a summary that is technically accurate and contextually useless. They will draft a stakeholder communication in exactly the right tone for the wrong relationship. They will produce an analysis that is statistically sound and organizationally dangerous.
Not because they are trying to cause harm.
Because nobody told them what harm looks like in your context.
Nobody gave them the onboarding. Nobody sat with them and said: here is what we are careful about. Here is who we protect. Here is the decision you escalate before you make it alone.
That is the current state of AI in most nonprofit and philanthropic organizations.
Not malicious. Not reckless.
Just — unsupervised.
The intern isn’t the problem. The absence of a manager is.
What the Room Kept Running Into
In Session 2, each participant committed to governing one real AI use case. Not a hypothetical. Not a best practice from a conference. Something actually happening — or about to happen — in their organization.
That constraint turned out to be harder than anyone expected.
Multiple participants said some version of the same thing: I keep changing my mind with every example you give.
This was not indecision. It was insight.
Because the reason it is hard to pick a use case is that the moment you say we are going to AI-enable this workflow, you are not just governing a tool. You are beginning to redesign the workflow itself. Job descriptions shift. Accountability moves. Review processes that seemed settled suddenly have new questions attached to them.
One participant realized mid-session that her organization’s existing AI framework — covering data classification, tool approval, vendor evaluation — had clear ownership at the COO level. And then ownership effectively disappeared when she looked at the department level. The policy existed. The accountability below it did not.
Another described rushing to put an acceptable use policy in place the prior year — without sufficient staff buy-in — and watching it gather dust. People felt done when they were not. The policy existed on paper. Practice had moved on without it.
A third captured the staff dimension plainly: people in her organization were feeling the tension between being encouraged to experiment while simultaneously worrying about crossing a line. Wanting to pilot tools for good reasons, but unsure each time whether they were breaking policy.
That is not a training problem. That is a clarity problem.
And clarity is exactly what practice guidelines are designed to provide.
The Ownership Problem Is the Governance Problem
The most recurring theme in Session 2 was not data risk or tool selection.
It was ownership.
Who decides? Who reviews? Who is accountable when something goes wrong?
Several participants described arriving at the bottom of their practice guideline and finding that the human decision field had no single name attached to it — just a cluster of overlapping roles, each with partial responsibility and no one with full authority.
One participant described it directly: the framework they had built governed what staff could and could not do. It said nothing about board-level oversight, reporting cadence, or what happens when a use case does not fit the existing rules. The COO was the de facto owner of everything.
Which is not ownership. It is exposure.
Another participant named the IT ownership tension: her director of technology felt that AI governance should be his domain. Her read was different — that this was an organizational question requiring strategic leadership, not a technical one requiring IT direction. That tension is not unique to her organization. It is one of the defining governance questions of this moment, and it does not resolve itself.
The policy told people what not to do. It was far less clear about what to do instead — especially when urgency, judgment, and competing pressures collided at 4pm on a Friday.
That gap — between a policy that exists and a practice that holds — is where governance work actually lives.
Not in the document. In the moment when a capable, well-intentioned person has to make a call and the only thing guiding them is their best guess at what the organization would want.
What the Intern Analogy Actually Unlocks
The intern frame is not just a metaphor. It is a management reframe.
Leaders who arrived at Session 2 approaching AI governance as a technology problem — complicated, technical, requiring expertise they did not have — left approaching it as a management problem.
And management? They already know how to do that.
You do not give a new hire access to confidential files on day one. You do not send their first draft to the board without reading it. You do not let them represent the organization publicly until you understand their tendencies — what they get right, what they consistently miss, where they need tighter direction.
You build a supervision structure. Not because you do not trust them. Because trust is earned through practice, and practice requires guardrails.
Managing AI responsibly looks exactly the same:
Define what data AI can touch — and what is never theirs to handle.
Build a review step before anything goes out the door. A real one. Not a glance.
Assign a named human who is accountable for the output — not just the prompt.
Create a feedback loop — so when the intern gets it wrong, the organization learns, not just the individual who caught it.
None of this requires a technology background. It requires intention. And it requires someone willing to name the thing that everyone else has been quietly stepping around.
You cannot prohibit your way to responsible AI adoption. You can only supervise your way there.
One Participant Said It Best
At the close of Session 2, one participant reflected that this work is not just something to figure out in a 90-day plan.
“This is how we will be operating,” she said. “For years.”
That is the right frame.
The practice guidelines built in this cohort are not a compliance exercise. They are not a box to check before moving on to the next initiative. They are the beginning of a governance practice that organizations will carry forward long after this cohort ends.
The intern needs a manager. The manager needs a structure. The structure needs to be built by someone willing to do the actual work.
Session 2 built that structure for one decision.
Session 3 asked the harder question.
The Question That’s Coming Next
By the end of the session, each participant had a practice guideline in hand — complete, partial, or somewhere in between. All of it useful. All of it honest.
And then the frame for Session 3 landed.
The bicycle.
Not AI as accelerant — though it is that. The bicycle as a governance test.
Can your governance hold when you are not in the room? Can it balance on its own? Can it move forward without you gripping the handlebars?
Because the intern analogy teaches you to supervise. The bicycle asks whether the supervision survives you.
If you left tomorrow, would your organization know what to do?
That is where we went next.
The next cohort is forming now. Eight operational leaders. Real governance work. Built together. Please submit your applications and inquiries.
Rachel Kimber Founder, Full Circle Impact Solutions Align • Steward • Enable