Generosity Is Not a Substitute for Investment

Welcome to data uncollected, a newsletter designed to enable nonprofits to listen, think, reflect, and talk about data we missed and are yet to collect. In this newsletter, we will talk about everything the raw data is capable of – from simple strategies of building equity into data+AI processes to how we can make a better community through purpose-driven analysis.

 ****************************************************************************

There is something troubling about watching AI support for nonprofits get translated into volunteer projects. I recently came out of a conversation with someone who spoke about a new offer on their platform: volunteering-based, tightly scoped AI support for nonprofits. The framing was efficiency and resource crunch — offering services like volunteers doing AI assessments in 2 hours, writing AI policies in 1 hour, and more — to help nonprofits move quickly with limited cost.

And that conversation made me uncomfortable. Today I want to co-explore, co-language-ize this discomfort.

Not because volunteers do not have much to offer, or because many of our nonprofits do not need help. Both are true. I volunteer 4-5 hours, 40-45 weeks out of 52 weeks for the past 8 years. Every bit of volunteering is a soul-balm to me.

And yet, the idea of offering two-hour volunteer support to get the AI assessment done made me uncomfortable.

Perhaps because the language of those volunteering opportunities highlighted a dangerous assumption: that some of the most consequential decisions a nonprofit can make about AI can be safely compressed into a donated hour, a packaged 45 to 60-minute working session, and a quick handoff.

Don’t get me wrong, the offer is generous for sure. A nonprofit gets help. A skilled volunteer shares expertise. An organization with limited resources gains access to something it might not otherwise be able to afford. In a sector where capacity is often underfunded, this can look like innovation.

But AI is not just another nice-to-have task to place on a volunteer marketplace. It's just not that kind of reality… AI policy writing, AI readiness assessments, internal use guidance, workflow redesign, and governance decisions are not minor back-office errands. They shape how an organization handles trust, risk, labor, data, decision-making, and community impact. They influence what staff are allowed to use, what donor or client information is processed, what biases are embedded, what shortcuts are normalized, and who is responsible when harm occurs.

Consider these examples:

• A domestic violence shelter is developing an AI intake screening process – it is making decisions about survivor safety and data confidentiality.

• A workforce development nonprofit deploying AI tools to match clients to job opportunities – it is making decisions that affect people’s economic futures.

• A youth-serving organization is building an AI chatbot for mental health navigation - it is operating in a territory where a poorly designed policy is way more than a minor inconvenience.

These are not some edge cases. They are the kinds of organizations most drawn to free support and the ones with the least margin for error.

That is why the instinct to support AI needs with volunteer-based activities deserves more scrutiny.

This is also why I keep coming back to a larger question now surfacing in philanthropy: can funders rise to the AI challenge in a way that matches the seriousness of this moment? A recent article from Inside Philanthropy – which I found via a post from Nathan Chappell (a leading voice on Responsible and Beneficial AI) - shares that AI needs to move to philanthropy’s front burner, and that while philanthropy cannot compete with technology companies in sheer dollars, it still has choices about where and how to exert influence. And extending that thought, I want to ask questions grounded in what that looks like on the nonprofit side. If foundations believe AI matters, then one of the clearest tests of that belief is whether they fund real nonprofit readiness — literacy, governance, implementation support, and accountable expertise — or whether they allow under-resourced organizations to meet this moment through donated fragments of labor.

Again, the question is not whether volunteers can be helpful. Of course they can. Skilled volunteers can offer meaningful support, especially at the early stages. They can introduce language, share examples, demonstrate tools, and help nonprofits think through possibilities. For some organizations, that first point of exposure may genuinely open a door (and I feel, as a life-long volunteer myself, I do not want any part of this essay to say volunteer doesn’t matter)

The problem begins when this kind of support is framed as enough.

Because once AI work is positioned as volunteer-appropriate infrastructure, we quietly teach the sector several dangerous lessons.

We teach nonprofits that AI governance is optional, lightweight, and low stakes. We teach funders that instead of investing real dollars in nonprofit readiness, they can route organizations toward donated labor. We teach intermediaries that access is the same thing as capacity. And we teach the market that important strategic work does not necessarily require paid, accountable expertise.

That is not a small shift. It changes expectations — and it changes what the sector believes it deserves. It also deepens inequity.

Well-resourced institutions will still pay for strong advisors, legal review, implementation support, security guidance, staff training, and change management. They will have people who can interrogate a draft policy, question a recommendation, and revise a roadmap. They will have time to pressure-test the work. A hospital system, a university, or a large social services agency navigating AI adoption will bring in specialized legal counsel, run internal pilots, and invest in staff change management. They will not be relying on a volunteer they met through an online platform.

Under-resourced nonprofits, meanwhile, may receive well-intentioned but uneven volunteer help. Some will get excellent volunteers. Others, well, it depends…Many will not know how to tell the difference — and that gap in organizational literacy is itself a product of underinvestment, not a reason to accept thinner support.

Let’s define the relationship between AI and the nonprofit sector: What Each Side Must Bring:

Before the sector can have an honest conversation about how AI should be resourced, it needs to be honest about what a real relationship between AI and the nonprofit world requires of both. This is not a one-way transaction. It should be an accountability-first symbiotic relationship.

What the nonprofit sector should demand of AI:

• Transparency about limitations. AI tools must be clear about what they cannot do, where they are likely to be wrong, and what human judgment they cannot replace — especially in high-stakes service delivery contexts.

• Equity by design, not as an afterthought. AI systems deployed in communities of color, low-income populations, or other marginalized groups must be developed and audited with those communities’ realities centered from the beginning, not patched for bias after harm has already occurred.

• Explainability that staff can actually use. Nonprofits should not have to accept “the model said so” as a sufficient answer. AI tools must be able to explain their outputs in language that non-technical staff can interrogate, challenge, and own.

• Data practices that honor constituent dignity. Any AI tool that touches client, donor, or community data must operate under data governance standards that treat the people behind that data as rights-holders, not inputs.

• Pricing and access models that do not exploit the mission. The nonprofit sector should not be forced to accept lower-quality tools, fewer features, or exploitative licensing terms simply because it operates on constrained budgets. Mission does not mean second-class access.

• Accountability when tools fail. AI vendors and platforms must have clear, enforceable mechanisms for redress when their tools cause harm in a nonprofit context — not just terms-of-service disclaimers that leave organizations absorbing all the risk.

• Long-term support, not abandonment. The sector should demand that AI tools offered to nonprofits come with genuine implementation support, updates, and continuity — not a launch and a handoff to communities that lack the internal capacity to maintain what they have been handed.

What AI — and those who develop, fund, and deploy it — should demand from the nonprofit sector:

• Organizational readiness before adoption. AI should not be implemented by organizations that have not done the foundational work of assessing their data practices, staff capacity, and governance structures. Adopting AI without readiness is not resourcefulness — it is risk transfer onto communities.

• Named accountability. Every AI tool in use at a nonprofit should have a person, team, or committee ( ideally backed by a policy document) — that makes both AI and the humans accountable for its use, monitoring, and termination if necessary.

• Honest engagement with bias and harm. The sector must be willing to ask uncomfortable questions about whether the tools it adopts reflect the values it claims, even when those questions slow down adoption or implicate vendor relationships.

• Investment in internal literacy. Nonprofits cannot outsource AI judgment entirely. The sector must invest in building the internal knowledge needed to evaluate recommendations, push back on vendors, and make independent decisions — even when that investment is hard to fund.

• Community voice in governance decisions. The people most affected by how AI is used in service delivery, advocacy, or resource allocation should have a meaningful role in shaping the governance of those tools. This is not optional participation — it is the sector’s core ethical obligation when demanding ethical technology.

• Willingness to say no. Not every AI application is appropriate for every organization or every context. The sector must develop the confidence to decline tools that do not serve its mission, even when they are free, popular, or funder-endorsed.

• Accountability to each other as humans of this sector. Nonprofits should share what they learn — what worked, what failed, what caused harm — rather than treating AI adoption as a competitive advantage or a reputational risk to manage quietly. Sector-wide learning requires sector-wide honesty. This understanding matters because the absence of mutual accountability is precisely what can make this a sustainable long-term conversation. When AI guidance arrives without any framework for what the nonprofit owes its constituents, and what AI owes the nonprofit, everyone gets to feel helpful while no one is actually responsible.

So, what can this sector do?

Let’s imagine and co-create our response.

When nonprofits need help with AI, the answer should not primarily be to send them to unpaid labor (or to one webinar or one conference). The answer should be for foundations and ecosystem funders to step up with money, advisory infrastructure, shared tools, trusted specialists, and learning spaces that are actually resourced.

If philanthropy believes AI matters — and, based on the flooding of articles, events, and tools in my inbox, it seems philanthropy does — then AI readiness should be treated as core nonprofit infrastructure, not a bonus service stitched together through goodwill (or when the board demands it one fine day).

I believe this can mean several possibilities for philanthropy to step up, such as

• Foundations subsidizing advisory hours with vetted specialists — not a one-size-fits-all online platform, but curated relationships with consultants who understand the regulatory environments, community obligations, and organizational realities of the nonprofits they serve.

• Pooled funding for sector-specific AI governance frameworks, developed with legal, equity, and implementation expertise baked in from the start.

• Coalition of funders investing together in a shared, adaptable AI policy library, for example, for social service organizations.

• Grants for staff learning and internal capacity — not one-day workshops, but multi-month cohorts where teams can learn together, adapt tools to their context, and build genuine institutional knowledge.

It also means honest grantmaking reform. As long as funders and foundations restrict overhead and refuse to fund strategic infrastructure, nonprofits will accept whatever free help is available — not because they want to, but because they have no other choice. Accountability for AI readiness must be matched by accountability in funding. Funders who believe AI governance is important need to show it in their grantmaking, not just in their convenings. Volunteers can still have a place (because I need my place as a volunteer in this ecosystem). There is room for orientation, peer learning, office hours, and bound technical help. But volunteering should be one layer in an ecosystem of support (while others do their part).

*************************************************************************************************

We often stop at the “AI is for efficiency” narrative. But it’s more than just about efficiency. It is about judgment. It is about justice. It is about who is asked for consent, who disappears within a dataset, who is expected to trust automated outputs, who is protected when tools fail, and who bears responsibility when decisions move far too quickly.

These are governance questions + leadership questions + equity questions.

And they deserve more than your and mine weekend afternoon.

What nonprofits need is not just access to AI help. They need durable, accountable, context-aware support that matches the seriousness of the decisions before them.

What funders and foundations need to recognize is that if they do not fund that support, they are not being neutral. They are actively shaping a system where the organizations with the least resources are asked to build their AI future on the thinnest possible scaffolding.

And I refuse to accept that as the best version of our possible generosity.

How we resource this moment will determine not just who adopts (AI) well, but who gets left holding the risk when adoption goes wrong.

*** So, what do I want from you today (my readers)?

Share with us: What feels true to your/your community’s reality here?