By Dimitry Dikman and Alex Spektor
Artificial Intelligence is no longer an “innovative add-on” in the nonprofit sector. It’s becoming an operating layer across everyday work: drafting donor communications, shaping grant narratives, summarizing reporting, and accelerating analysis. That shift delivers real productivity gains — and it also introduces a predictable category of failures.
The biggest problems are rarely technical. They’re governance problems: unclear approvals, weak quality controls, sloppy data boundaries, and overconfident public claims. In other words, the constraint isn’t which model or tool an organization uses. The constraint is whether decision-making, accountability, and verification can keep up with the new speed.
This article frames AI as infrastructure and explains why governance is now the limiting factor. It then uses fiscal sponsorship as a high-signal case — a structure that makes governance questions unavoidable — and closes with what funders are likely to start asking as part of AI-era due diligence.
A few years ago, AI use in nonprofits tended to be occasional and isolated: a draft here, a brainstorm there. Today, it’s increasingly embedded in workflows. What makes that important isn’t novelty — it’s scale. When output volume rises and cycle times shrink, small errors compound quickly. A single sloppy sentence can travel from an internal draft to a donor email, a grant report, a board memo, and a public post within hours.
This “infrastructure” shift is visible in how nonprofits report using AI. In a 2025 survey by the Center for Effective Philanthropy (CEP), nonprofits most commonly reported using AI for communications (84%), internal productivity (63%), and development/fundraising (61%) — exactly the functions where language, accuracy, and credibility matter most.
CEP also notes that leaders describe AI as helpful for drafting emails, policies, procedures, meeting summaries, and other documents — and that many nonprofits use it for development work, including drafting or refining grant applications and grant reports.
The practical implication is straightforward: when AI becomes a normal part of producing donor- and funder-facing content, the question is no longer “Should we use AI?” The question is: What happens to decision quality when content production accelerates?
The most useful way to think about AI risk in nonprofits is governance risk. The core dangers are not “the tool is powerful,” but that AI makes it easy to move faster than the organization’s controls — and to sound confident even when the underlying claims are weak.
Governance in an AI-enabled environment can be simplified into three levers:
Decision rights: Who approves what — and at what risk level?
Controls: What must be reviewed, verified, and documented before it goes out?
Data discipline: What information can be used, where, and under what boundaries?
These levers matter because confidence is not the same as accuracy. Research from Stanford HAI on legal AI tools found hallucinations in roughly one out of six benchmarking queries — a clean reminder that even strong systems can produce plausible but wrong outputs in high-stakes contexts. That doesn’t mean “don’t use AI.” It means verification controls must be intentional and proportional to risk.
CEP’s 2025 findings also suggest this is already a mainstream leadership issue. Foundations and nonprofits reported concerns related to accuracy, bias, staff expertise, and security — concerns that map directly to governance capacity (who knows what, who reviews what, and what guardrails exist).
Finally, governance matters more now because credibility is more fragile. The 2025 Edelman Trust Barometer frames misinformation and credibility as ongoing anxieties in the broader environment. In a low-trust context, “overconfident claims” carry a larger reputational downside — especially when AI accelerates publishing.
If governance is the bottleneck, fiscal sponsorship is where that bottleneck becomes impossible to ignore.
Fiscal sponsorship is often discussed as an administrative structure. But it is also a governance structure — one where oversight and fiduciary responsibility are explicit. That clarity is exactly why fiscal sponsorship is a high-signal lens for AI governance.
When AI is used inside a fiscally sponsored project, the first things that tend to go wrong are not “technical.” They are governance failures that show up in predictable places:
Donor-facing commitments that quietly overpromise (“all funds will be used exclusively for…”).
Restricted-funds language that is too rigid, too vague, or inconsistent across documents.
Grant compliance commitments that drift from what is actually feasible to document.
Reputational exposure when a public claim turns out to be unverified or misstated.
Funders already know this terrain. The National Network of Fiscal Sponsors’ guidance for funders emphasizes sponsor oversight and warns against “pass-through” arrangements — a diligence expectation that becomes even more salient when AI is involved in drafting and reporting.
The restricted-funds issue is particularly instructive. Legal guidance on restricted or designated gifts underscores that restrictions can create legal and reputational consequences if donor intent isn’t honored — which makes AI-generated language that “sounds right” but isn’t precise a real governance risk.
The lesson isn’t that fiscally sponsored projects are uniquely vulnerable. The lesson is that fiscal sponsorship makes accountability visible: it forces clarity about who approves donor-facing language, how restrictions are recorded, and what documentation is required to substantiate use of funds.
As AI becomes embedded in nonprofit operations, due diligence is likely to evolve. The point will be whether the organization can demonstrate verification discipline, clear accountability, and credible controls for donor- and funder-facing claims.
In practice, funders will start probing for “maturity signals” — not policies for their own sake, but evidence that governance can keep pace with speed. Expect questions like:
What counts as “verified” when AI is involved — especially in impact statements, citations, and public claims?
Who owns accountability for AI-shaped outputs (a named decision owner, not a vague “the team”)?
What will the organization not automate (clear red lines for high-stakes communications)?
How do corrections happen when something is wrong (donor communications, public updates, documentation)?
For fiscally sponsored projects: when AI is used, who reviews and approves what goes public — and who is responsible for keeping sensitive data out of AI tools?
CEP’s 2025 research supports the premise that this is becoming part of the mainstream risk conversation: leaders in both foundations and nonprofits cite concerns that translate directly into governance diligence — accuracy, bias, staff capacity, and security.
For fiscally sponsored projects, the diligence conversation is even more legible because funder-facing guidance already emphasizes sponsor oversight and cautions against pass-through dynamics. In an AI-enabled environment, that same logic expands naturally to AI-shaped outputs: who reviewed them, what was verified, and where accountability sits.
AI increases speed. Governance determines whether speed produces trust or risk. Fiscal sponsorship is a useful lens because it forces clarity on oversight and accountability — the two things AI adoption most often exposes.
If your project is using AI in donor- or funder-facing work — especially under fiscal sponsorship — governance is what protects trust while preserving speed.
Learn more about fiscal sponsorship, or apply for fiscal sponsorship through our short application form.
About the Authors
Dimitry Dikman is the founder of Group 36, a Pennsylvania-based 501(c)(3) providing fiscal sponsorship and grant administration. He entered the nonprofit sector after years in business and consulting and has spent more than a decade helping leaders strengthen strategy, budgeting, and governance. His approach emphasizes fiduciary responsibility and practical controls that keep charitable funds aligned with donor intent and funder expectations.
Alex Spektor is a technology leader and platform architect with decades of experience building dependable software in regulated environments. His recent work centers on applied AI, particularly agentic systems. His work emphasizes the unglamorous parts that make AI usable in real life: clear tool boundaries, guardrails, and evaluation practices that keep agent behavior reliable over time.
Sources & Further Reading
Center for Effective Philanthropy (CEP) — How Nonprofits and Foundations Are Using AI (2025)
Stanford HAI — AI on Trial: Legal Models Hallucinate in Benchmarking (2025)
Edelman — Trust Barometer (2025)
National Network of Fiscal Sponsors (FiscalSponsors.org) — 10 Questions Funders Should Ask Grantees Who Use Fiscal Sponsors