Article 6 — Governance in the Wild
Detroit in early spring feels like a city mid-sentence.
Construction cranes everywhere. Streets partially blocked. Buildings being restored next to others still waiting their turn.
It feels uncertain.
Like a place still deciding what it will become.
Which, in hindsight, was a fitting place to spend a week talking about artificial intelligence.
The morning of the opening keynote at the Nonprofit Technology Conference, Google Maps sent me down a street behind the convention center that didn’t feel like a street I was supposed to be on.
It was raining just enough to get me wet.
I ended up jogging the last block, arriving slightly late, slipping into a table, asking someone for the wifi password, and opening my laptop while I dried out with a cup of coffee.
On stage, technologist and writer Anil Dash was talking about power.
Not abstract power.
Institutional power.
One idea stuck with me: the higher you move in leadership, the more your job becomes navigating difficult tradeoffs.
That observation lands differently in a room full of nonprofit leaders.
Because in this sector, leadership rarely means unilateral authority.
It means navigating tradeoffs in public.
Mission and money. Speed and stewardship. Innovation and trust.
Those tensions would show up very clearly later that afternoon.
The Debate
My role in the conference debate was not planned months in advance.
On Sunday night, Ben Childers reached out. Ben — who also organized a debate I participated in at the Fundraising.AI Summit last fall — had a last-minute opening. Anne Murphy, who had originally been scheduled to participate, had a conflict. Could I step in?
I had about two days to prepare.
I was juggling fractional CIO work, wrapping up responsibilities before leaving for the conference, and trying to mentally transition into travel mode.
So my internal response looked something like this:
Yes.
No.
Yes.
Anyone who has tried to read their calendar on a Sunday night understands how this goes.
Eventually the answer became yes.
Standing Room Only
The debate itself was scheduled in a room down a long anonymous hallway in the convention center.
Nothing about the location suggested it would become one of the most talked-about sessions of the week.
But by the time the session started the room was full.
People were standing along the walls.
Later I learned people had started texting colleagues once the debate really got going — telling them to get there before the closing arguments.
The format was Oxford-style.
Two teams. Opening arguments. Rebuttals. Audience vote.
The motion:
Nonprofits that fail to adopt generative AI will be unable to compete for talent and funding within five years.
Arguing for the motion: Cheryl Contee, CIO at BrightWorks AI, and Ben Miller, SVP of Data Science and Analytics at Bonterra. Their core claim: AI multiplies organizational capacity, and refusing to adopt it responsibly means choosing higher cost and lower impact.
Arguing against: Craig Johnson, whose position centered on governance and trust as the real competitive advantages — and me.
Our argument was simple: AI is not a competitive advantage. It is a highlighter — one that reveals data chaos, process problems, and organizational weak points that already exist. The real risk facing nonprofits is not failing to adopt AI. It’s bad adoption.
The room was tense.
The topic was controversial.
But the real tension came from something else.
The people debating weren’t strangers.
They were colleagues.
Collaborators.
People we respect.
People we learn from.
People we’d see again the next morning in the hallway.
Which made the disagreement feel real.
So at one point I paused and said something simple.
“Hey everyone. We’re friends. We like each other. We work together. But this is intense. Thank you for going on this journey with us.”
The reaction was immediate.
Nods.
Smiles.
And applause.
You could feel the room relax.
Everyone had been sitting in the tension we had deliberately created. Acknowledging it made it easier for people to stay present.
People could vote honestly.
They could disagree.
They could stay curious.
That moment changed the tone of the entire session.
Psychological Safety in Action
What happened in that room is something organizational researcher Amy Edmondson calls psychological safety.
Psychological safety isn’t about avoiding difficult conversations.
It’s about making those conversations possible.
Ruchika Tulshyan writes in Inclusion on Purpose that inclusion requires leaders to deliberately create environments where disagreement can exist without social penalty.
It sounds simple. It rarely is.
Consider what Dr. Joy Buolamwini discovered while building a program to project digital masks onto faces at MIT. The facial recognition software she was using couldn’t detect her face at all — until she put on a white mask. The training data hadn’t included enough faces that looked like hers. One biased algorithm, built by a homogeneous team that never noticed the gap, traveled from lab to lab across multiple countries before anyone caught it.
That’s what ungoverned technology does.
It doesn’t just reflect existing inequity. It replicates it — rapidly, and at scale. Healthcare algorithms have been shown to recommend additional care for white patients while passing over equally sick Black patients. Not because anyone designed them to discriminate. Because no one designed them carefully enough not to.
Dr. Buolamwini went on to found the Algorithmic Justice League and her story is documented in the Netflix film Coded Bias — required watching for anyone making decisions about AI in their organization.
This is what I mean when I say the real risk is bad adoption.
AI doesn’t create new problems.
It accelerates the ones we already have — and makes them harder to see, harder to trace, and harder to undo.
That’s why the conversation in Detroit mattered.
We created tension.
Then we created enough safety for people to stay with it.
The result was one of the most honest conversations about AI I’ve seen in the sector so far.
The Result
The audience voted before and after the debate.
By the end, 60% rejected the motion. Only 38% approved.
That’s not a close result. It’s a decisive one.
Not because people oppose AI.
But because many in the room recognized something deeper.
The risk facing nonprofits is not failing to adopt AI.
The risk is bad adoption.
Sloppy processes accelerated.
Harmful workflows scaled.
Relationships and trust damaged by systems implemented too quickly.
That concern surfaced again and again throughout the conference.
And it points to a larger pattern.
Societies often adopt powerful technologies long before they develop the norms and rules needed to govern them.
Which brings us to another technology that once spread just as quickly: the automobile.
AI and the Automobile Era
I’ve written before about how the automobile era mirrors AI’s current moment. That comparison keeps feeling relevant — so I want to return to it here.
In the early 1900s, cars spread rapidly across American roads.
But there were almost no rules.
No standardized signals. No driver licensing. No clear right-of-way.
The results were predictable.
Crashes increased. Pedestrians were injured. Public concern grew.
In 1901, Connecticut adopted the first speed-limit law in the United States — an early attempt to manage the risks of this new technology.
Over time, additional standards emerged: traffic lights, driver licensing, and eventually the now-familiar octagonal stop sign.
These rules didn’t stop innovation.
They made widespread adoption possible.
AI today is following a similar trajectory.
The technology is already embedded in hiring, lending, healthcare, education, and government systems.
At the same time, researchers are documenting evidence of bias, privacy violations, and concentrated technological power.
History suggests something important:
innovation eventually demands governance.
The Governance Question
The Detroit debate revealed something important.
The nonprofit sector is not divided into “pro-AI” and “anti-AI.”
What we are navigating are tradeoffs.
Speed and reflection.
Innovation and stewardship.
Efficiency and equity.
These are not problems to solve once.
They are tensions to manage.
Leadership thinker Barry Johnson describes these dynamics as polarities rather than problems.
Both sides contain value.
Both sides contain risk.
Move too quickly and organizations risk accelerating harmful systems.
Move too slowly and organizations miss opportunities to reduce administrative burden and learn faster.
Governance helps institutions make these tensions explicit and navigate them responsibly.
Not eliminate them.
Leadership in the AI Era
The Detroit debate made something clear.
Technology alone will not determine how AI shapes the nonprofit sector.
Leadership will.
Leaders will decide:
What systems we trust.
What risks we accept.
What values guide the design of our tools.
And how we ensure that technology serves communities rather than extracting from them.
Those are governance questions.
And they don’t belong to one side of the sector.
Nonprofits and philanthropy are navigating them together.
Because the systems shaping our work belong to both.
What Comes Next
Detroit clarified something for me.
The governance questions facing nonprofits and philanthropy aren’t separate problems to be solved by individual organizations.
They are shared problems.
And the norms we establish now — who holds decision-making power over these systems, who is included, who is listened to — will shape how AI moves through this sector for years to come.
A cohort alone won’t solve that. But it’s a room where the right people can begin to build those norms together — in relationship, not in isolation, and not handed down from above.
That’s why this spring I’m convening a small cohort for COOs and senior operational leaders in nonprofit and philanthropic organizations, presented in collaboration with Philanthropy New York and Community Resource Exchange.
Two seats are already filled. The cohort is capped at 8 leaders and opens in April 2026.
Closing
Detroit was still mid-sentence when I left.
The cranes were still up. The streets still blocked. The city still deciding.
I think that’s where we are with AI too.
Not at the beginning. Not at the end. Somewhere in the middle of a sentence that the sector is still writing.
What happens next won’t be determined by the tools.
It will be determined by the people in rooms like that one — willing to sit in tension, stay curious, and take responsibility for what we build together.
Resources
Amy Edmondson — Psychological Safety research hbs.edu/faculty/Pages/profile.aspx?facId=6451
Ruchika Tulshyan — Inclusion on Purpose
Dr. Joy Buolamwini — Algorithmic Justice League ajl.org
Coded Bias — documentary on Netflix netflix.com/title/81328723
Barry Johnson — Polarity Management framework polaritypartnerships.com
AI Now Institute — Research on AI harms and govern