What AI Governance Actually Looks Like
This week I’m heading to the Nonprofit Technology Conference (NTEN), where I serve as NTEN's Artificial Intelligence nonprofit community group co-lead.
Every year the conversation about technology evolves.
This year the shift is unmistakable.
Everyone is experimenting with AI.
Very few organizations are governing it.
Over the past several articles in this series, I’ve argued three things:
AI is a leadership stress test.
Acceleration exposes weak governance.
Responsibility cannot be abdicated.
The next question leaders ask is the right one:
What does responsible AI actually look like inside an organization?
Not in theory.
Not in a white paper.
Inside real workflows, with real staff, making real decisions.
This article begins to answer that question.
Efficiency Is Where Everyone Starts
Most AI conversations begin with efficiency.
Faster meeting notes.
Faster summaries.
Faster reports.
Those use cases are understandable. Nonprofits and foundations operate with limited capacity, and AI tools promise relief.
But efficiency is the narrowest lens we can use.
Throughout this series I’ve framed AI decision-making through the E4 lens:
Efficient
Effective
Expansive
Equitable
Efficiency asks: Can we do this faster?
Effectiveness asks: Does this actually improve outcomes?
Expansiveness asks: Who else gains access when we redesign the system?
Equity asks: Who benefits — and who might be excluded?
If we only optimize for efficiency, we risk accelerating the very dynamics philanthropy exists to address.
Speed can land you in the wrong place faster.
Which is why readiness is not simply about tools.
It is about governance.
The Moment When Policy Meets Reality
Recently I worked alongside Valarie Ehrlich of Mission Bloom on an engagement with a regional family foundation that had already done something most institutions haven’t.
They had written an AI policy.
It was thoughtful. Values-driven. Grounded in commitments to transparency, human accountability, and grantee trust.
But leadership understood something important:
Policy alone does not change behavior.
Staff still needed clarity about the questions that arise in everyday work:
Can we use AI to summarize grantee reports?
What tools are appropriate for internal meetings?
What happens if an AI summary misrepresents a proposal?
When should AI use be disclosed?
How do we protect partner data?
So instead of treating the policy as the finish line, we treated it as the starting point.
A Pressure Test, Not a Compliance Test
Our goal was to translate principles into practice.
We began with a working session with the foundation’s AI governance committee — a group of staff who had invested significant effort into building the policy and now needed to stress-test it against reality.
The framing mattered enormously.
We were explicit from the start:
This is not a compliance audit.
The goal isn’t to find violations.
The goal is to find ambiguity — because ambiguity is where governance breaks down.
Responsible people can still make risky decisions when expectations are unclear.
We introduced realistic scenarios drawn from actual nonprofit workflows.
Each one was deliberately messy — the kind of situation a thoughtful, busy professional might actually face.
A program officer, racing to meet a board deadline, pastes excerpts from confidential grantee reports into a free AI tool to speed up analysis. She anonymizes names but includes program data and outcomes. The summary is excellent and ends up shaping a board slide.
A staff member uses an approved AI tool to draft a research summary, reviews it quickly, and sends it to leadership. Days later a partner points out a key statistic was subtly misquoted.
A staffer runs an AI analysis of a grantee’s public materials ahead of a site visit. The tool flags what looks like financial strain based on tone and phrasing — and that interpretation quietly shapes the questions asked in the meeting.
In each case: no bad actors.
Just thoughtful professionals working under time pressure in the absence of clear guidance.
The exercise surfaced something the committee felt immediately.
The policy told people what not to do.
It was far less clear about what to do instead.
That gap is where governance work actually lives.
From Ambiguity to Guidance
Once those gray areas became visible, we began building practical guidance together.
One of the first outputs was a practice guideline for AI meeting notetakers — one of the most common AI use cases across the sector, and one that raises more governance questions than it initially appears to.
Who consents to recording?
Where are transcripts stored?
Who verifies accuracy?
How long are records retained?
How do we ensure an AI summary doesn’t quietly become the official record?
The guideline the committee developed didn’t just answer those questions.
It built a workflow.
AI generates the initial notes.
A human validates for accuracy, context, and tone.
Only then is the output shared — with clear disclosure language staff can use.
What made it work wasn’t the document.
It was the process of building it together.
The committee arrived as policy stewards.
They left as practice designers.
That shift — from compliance managers to governance architects — is what responsible AI adoption actually requires.
What Changes When Governance Works
Something interesting happens when organizations move from theory to practice.
Fear decreases.
Staff who were initially cautious about AI experimentation become more confident because they understand the boundaries.
They stop asking:
Are we allowed to use this?
And start asking:
How can we use this responsibly?
That shift marks the beginning of institutional learning.
It also marks the beginning of something strategically valuable:
A culture where governance is experienced not as restriction — but as infrastructure.
Where policy is not a static document in a shared drive but a living framework staff actively improve.
That is what readiness looks like in practice.
Governance Is Now a Board-Level Obligation
AI governance is no longer simply a technical issue.
It is tied directly to the governance responsibilities boards already hold:
Duty of care
Duty of loyalty
Duty of obedience
These obligations now have a technology dimension many boards have not fully caught up to yet.
Leadership must be able to answer:
What AI tools are staff using?
What data is entering those systems?
What decisions are being influenced by AI outputs?
Who is accountable for those outcomes?
These questions are not going away.
They are becoming part of everyday organizational governance.
The Methodology: Align • Steward • Enable
Through engagements like this one, a consistent leadership pattern has emerged.
Align
Leadership clarifies goals, risk tolerance, and decision ownership before tools are selected.
Governance cannot succeed if leaders disagree about what responsible AI use looks like.
Steward
Organizations develop guardrails that protect mission, community trust, and institutional integrity.
Policy and scenario pressure-testing happen here.
Enable
Staff gain the tools, guidance, and confidence to experiment responsibly.
Innovation becomes possible because the boundaries are clear.
This progression moves organizations from uncertainty to intentional adoption.
Philanthropy’s Obligation
For-profit technology companies are driving AI development at extraordinary speed.
Their incentives are clear:
Scale.
Market share.
Return on investment.
The social sector operates under a different mandate.
Our responsibility is not simply to adopt technology.
It is to ensure technology strengthens communities rather than extracting from them.
That includes protecting against:
Data extraction from vulnerable populations
Opaque algorithmic decision-making
Systems that quietly reinforce structural inequities
Philanthropy holds a uniquely powerful position.
It has resources.
It has independence.
And it has the ability to shape norms before they become defaults.
The MacArthur Foundation has published its AI policy publicly as a field resource.
The Rockefeller Foundation’s AI for Development initiative centers equity in AI design.
These are early models.
But models matter only if institutions are willing to do the internal work of translating principle into practice.
Responsible AI governance is not a technical project.
It is leadership work.
The Question That Remains
AI is already inside our workflows.
Staff are using it to research, summarize, analyze, and generate content.
The question leaders face is no longer whether AI will enter their organizations.
It already has.
The question is whether we will govern its use intentionally — or allow it to shape decisions invisibly.
The organizations that begin building these capabilities now will not only reduce risk.
They will position themselves to accelerate impact.
From Diagnostic to Direction
If your organization is somewhere between curiosity and readiness — if you have staff experimenting, board members asking questions, and no shared governance architecture yet — you are not behind.
You are early.
But early without structure becomes risky faster than it feels.
That is why we are launching the AI Governance & Operational Leadership Cohort through Full Circle Impact Solutions.
Designed for COOs, executive leaders, and board-facing operators navigating:
compliance risk
board expectations
operational modernization
staff experimentation
This is not AI training.
It is governance architecture for institutions that intend to lead.
Participants leave with:
mapped AI decision ownership
a defined disclosure posture
drafted governance guardrails
pressure-tested scenarios
a 90-day implementation roadmap
a private strategy session for organization-specific challenges
Philanthropy seats underwrite nonprofit participation to ensure power-balanced learning.
Cohort capped at 8 leaders.
Launch: April 2026.
If this is the work your institution needs,
let’s talk.
Align • Steward • Enable
Policy without practice is intention without impact.
Responsible AI governance is not about slowing innovation. It is about ensuring innovation strengthens the communities philanthropy exists to serve.
The organizations doing this well are not the ones adopting tools the fastest. They are the ones building the strongest governance infrastructure around them.
That work begins with alignment.
It continues with stewardship.
And it succeeds when organizations enable staff to experiment responsibly.
The institutions that close this gap now will not only reduce risk.
They will shape how this technology serves the public good.
In the next piece: where AI can expand philanthropy’s mission — not just accelerate its operations. What does it look like when AI redistributes insight instead of concentrating it?
Sources & Further Reading
Mission Bloom — https://www.missionbloom.org
MacArthur Foundation Artificial Intelligence Policy
https://www.macfound.org/about/our-policies/artificial-intelligence
Rockefeller Foundation AI for Development
https://www.rockefellerfoundation.org/initiatives/ai-for-development
Center for Effective Philanthropy — Responsible AI Guidance
https://cep.org/blog/responsible-ai-how-philanthropy-can-and-should-support-the-movement/
NTEN — AI for Nonprofits Resource Hub https://www.nten.org/learn/resource-hubs/artificial-intelligence
NTEN Nonprofit Technology Conference
https://www.nten.org/gather/ntc/program
Kimber, Ehrlich, Vance & Osei (2025)
Collective Learning in Philanthropy — The Foundation Review
https://scholarworks.gvsu.edu/tfr/vol17/iss3/7/
Kimber & Vance — Purpose as a Compass, Philanthropy New York
https://philanthropynewyork.org/news/purpose-compass-grant-reporting-what-purpose-linking-values-vision-and-practice
Kimber — From Stuck to Strategic: Leading AI Team Alignment
https://www.resilia.com/webinar/from-stuck-to-strategic
Series Article 4 — From Stuck to Strategic
https://rachelkimber.substack.com/p/from-stuck-to-strategic