In the last piece, I argued that AI is a leadership stress test.
It can hold the whole essay question.
But only we can answer the why.
What I didn’t say — at least not directly — is this:
AI is not arriving into calm conditions.
It’s landing in systems where some absorb shock through reserves, and others absorb it through exhaustion.
That difference matters.
Because acceleration does not distribute consequences evenly.
Acceleration Is Not Theoretical
You don’t need a futurist to tell you the pace has changed.
Climate disruption is no longer abstract. It is geographically uneven, socially stratified, and accelerating faster than most institutional response cycles. The IPCC’s adaptation reports make clear that resilience correlates strongly with access to capital, governance capacity, and institutional infrastructure.¹
Some communities have cushion.
Others do not.
Institutional lag isn’t hypothetical. It’s operational.
We see it in disaster recovery timelines.
In funding cycles that can’t adapt midstream.
In reporting structures that assume stability where volatility is now the norm.
Acceleration is already here.
AI doesn’t create that pressure.
It multiplies it.
The Stanford AI Index documents exponential growth in model capability and deployment speed across sectors.³ Decision cycles are compressing. Deployment windows are shrinking. Institutional deliberation remains slow.
And when speed outpaces structure, something breaks.
But we also know what happens when infrastructure is built in advance.
June 12, 2016. Pulse nightclub. Orlando.
Within days, staff at Arcus Foundation were already connected to people on the ground. Nonprofits were prepared to receive funds. Families needed immediate support.
Standard protocol would have required board approval. Those meetings only happen quarterly.
Instead, the Executive Director and board developed a plan to activate resources. Money moved in less than two weeks. Arcus committed $100,000 directly and helped mobilize more than $1.5 million in national funding commitments.²
It worked because relationships were already in place. Role clarity existed before the crisis. Values were explicit. Trust allowed speed without chaos.
After Pulse, Arcus formalized what they had done instinctively: delegated authority grants.
That’s not a story about speed.
It’s a story about what has to be true before speed is safe.
AI accelerates decisions.
Governance determines whether those decisions stabilize or destabilize institutions.
The Risk Is Not Automation. It’s Asymmetry.
The public conversation about AI focuses on automation.
Jobs. Tools. Efficiency.
But that’s not the core risk in philanthropy.
The risk is uneven preparedness.
As Kate Crawford argues in Atlas of AI, technological systems inherit and amplify existing power structures.⁴ Acceleration without equitable infrastructure does not level the field. It sharpens it.
Wealth absorbs volatility differently than under-resourced institutions.
Large foundations can:
Hire advisors
Pilot cautiously
Establish oversight committees
Pause and recalibrate
Smaller institutions often cannot.
They adopt what they must.
React instead of design.
Absorb risk instead of allocate it.
We’ve seen this before in grant reporting systems that require nonprofits to reformat identical information across multiple platforms—serving funder reporting needs more than community ones.⁵
Those systems are rigid.
But they are predictable.
Current systems are rigid and predictable.
AI systems are rigid and opaque.
Opacity changes the risk profile.
Now imagine layering AI on top.
A funder deploys a model to screen applications. The nonprofit writes in narrative form, centering community voice and relational outcomes. But the model was trained on historical funding patterns emphasizing quantifiable metrics and institutional scale.
The proposal is flagged as “misaligned.”
Not because the work lacks merit.
Because it isn’t legible to the model.
The nonprofit never sees the scoring logic.
Never sees the weighting.
Never sees what was omitted.
They simply aren’t funded.
That’s not neutral automation.
That’s structural coercion.
And it mirrors broader philanthropic asymmetries documented by CEP — where capacity shapes influence, and risk is often externalized downstream.⁶
AI layered onto imbalance does not flatten hierarchy.
It operationalizes it.
When Acceleration Meets Governance Gaps
Consider what happens when AI tools move faster than governance structures can respond:
A model flags organizations as “high risk,” and no one reviews the training assumptions.
Automated summaries reshape how board members interpret proposals.
Efficiency dashboards crowd out relational indicators because they are easier to quantify.
None of this requires malice.
It only requires speed.
I’ve been in board meetings where a single question shifted a decision:
“But what is the actual need here?”
That question depends on someone recognizing what the summary left out.
If the summary is AI-generated, and no one understands what was emphasized or omitted, the corrective moment may never occur.
And here is the executive risk:
When AI embeds flawed assumptions into decision systems, harm scales invisibly.
Not as scandal.
As drift.
Reputational erosion.
Community distrust.
Misalignment between stated values and actual allocation patterns.
Responsible AI frameworks emphasize governance before deployment.⁷ Yet sector research suggests adoption often outpaces oversight.⁸
When acceleration meets weak governance, asymmetry compounds quietly.
Philanthropy’s Structural Position
This is the uncomfortable part.
Philanthropy sits inside concentrated capital.
It funds resilience.
It experiments with innovation.
It often benefits from insulation its grantees do not share.
CEP has documented persistent gaps between funder perception and grantee experience.⁹ The issue is not intent. It is structure.
If volatility is privately buffered and publicly absorbed, instability compounds.
If AI increases decision speed without increasing decision scrutiny, hierarchy hardens.
I have received the call that comes after the decision — when someone realizes they were never part of the conversation shaping their reality.
That is what happens when speed outpaces relationship.
AI, if misapplied, makes that dynamic invisible.
Because the system feels efficient.
It just no longer feels accountable.
And legitimacy, once eroded, is difficult to restore.
Stability Lives in Governance
When systems accelerate, stability does not live in prediction.
It lives in governance.
In:
Clear decision rights
Explicit value articulation
Defined escalation pathways
Transparent review mechanisms
Role clarity
Institutional memory
I have seen well-resourced institutions falter because no one knew who was accountable for what.
And I have seen smaller organizations endure complexity because governance clarity allowed them to move decisively.
Scale is not the determining variable.
Governance clarity is.
AI is not destabilizing because it is intelligent.
It is destabilizing because it compresses time.
Compressed time magnifies weak governance.
Strong governance absorbs acceleration.
Weak governance exports it.
Preparedness Is a Strategic Choice
The question is not whether AI will change philanthropy.
It already is.
The question is whether leadership treats acceleration as procurement — or governance.
Tools can be purchased quickly.
Capacity cannot.
If we misdiagnose AI as a technology upgrade instead of a governance shift, we will build the wrong safeguards.
And when the consequences surface, they will not look like technical errors.
They will look like broken trust.
If we build infrastructure first, AI becomes something different:
Not a destabilizer.
A coordination tool.
Not a hierarchy amplifier.
A resilience multiplier.
But only if governance precedes deployment.
And yet.
Even strong governance assumes we understand the technology we are governing.
In the next piece, I will explore what makes AI categorically different from previous technologies — and why misunderstanding that distinction leads institutions to prepare for the wrong risks.
Because if we treat AI like past technology, we will misallocate the response.
And under acceleration, misallocation compounds quickly.
More on that next.
References & Further Reading
Intergovernmental Panel on Climate Change (IPCC). Climate Change 2023: Impacts, Adaptation and Vulnerability.
https://www.ipcc.ch/report/ar6/wg2/Rizzo, C. T. (2018). Crisis Philanthropy: Two Responses to the Pulse Tragedy in Orlando. The Foundation Review, 10(1).
https://doi.org/10.9707/1944-5660.1406Stanford Institute for Human-Centered Artificial Intelligence (HAI). AI Index Report 2024.
https://aiindex.stanford.edu/report/Crawford, Kate. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence.
https://yalebooks.yale.edu/book/9780300209570/atlas-of-ai/Kimber, R., Ehrlich, V., Vance, B., & Osei, A. (2025). Collective Learning in Philanthropy: AI, Trust, and the Future of Grant Reporting. The Foundation Review (forthcoming).
Center for Effective Philanthropy (CEP). Bridging the Gap: Grantee Perspectives on Intermediaries.
https://cep.org/portfolio/bridging-the-gap/Partnership on AI. Responsible AI Practices. https://partnershiponai.org/
Center for Effective Philanthropy (CEP). The Case for 21st Century Philanthropy.
https://cep.org/portfolio/the-case-for-21st-century-philanthropy/Center for Effective Philanthropy (CEP). Listening to Those We Serve.
https://cep.org/portfolio/listening-to-those-we-serve/Government of Australia. Falepili Union Treaty Between Australia and Tuvalu (2023).
https://www.dfat.gov.au/geo/tuvalu/falepili-union-treatyBBC News. Australia Offers Climate Visa Pathway to Tuvalu Residents as Sea Levels Rise.
https://www.bbc.com/news/world-australia-67426632
If you’re a foundation leader preparing for AI adoption, I work with institutions to build governance infrastructure before deploying tools—ensuring acceleration strengthens relationships instead of straining them.
Knowledge gaps being exacerbated through tech are a well-studied phenomenon, but I would also like to consider the possibilities if AI *doesn't* live up to expectations. A lot of upheaval is being forced upon many aspects of life through the insistence that AI will change everything. And those who don't wish to participate in this way are being told that they'll forever remain in a societal underclass. The powers-that-be almost NEED this AI upheaval to happen.
Entire labour markets are expected to be restructured on the promise of AI. If the hype is to be believed, entry-level jobs in most fields will be obsolete before the decade is out. This is while promising that the best workers will be more productive than ever before thanks to AI. This is killing the goose that lays the golden egg, since every "top performer" was once an entry level participant. Without entry-level participants who will one day become senior, who trains the next-gen? The situation is even worse if AI does not meet the promise, because we've destroyed established pathways in every industry, only to rebuild them in a world with fewer worker protections than 100 years ago.
Calling for 'governance' and 'stability' rings hollow without a direct focus on the human impact. It's assuming a neutral playing field of rational actors, ignoring the active efforts that have been made to minimise labour's share of the income. All the 'productivity' gains of the implementation of AI has gone to the boardroom, taking away from the workers. And these gains are happening at the expense of the quality of work. We all know we live in a slop-driven dead internet. It's a shocking decline in craftsmanship thanks to the sidelining of coders, writers, artists and personal assistants through LLMs, and it will only get more pronounced.
The actual political questions that governance should be asking are_
- who owns the productivity gains?
- who bears the displacement costs?
- how does AI actually contribute to lifting the floor of society?