I have been thinking about where donors might make some of the most useful investments in AI right now, and I keep coming back to something that is not particularly new or shiny: nonprofit infrastructure.

AI is arriving in a sector that has spent decades underinvesting in the systems underneath the work. That creates a problem, but also an opportunity.

For donors, foundations, and the advisors helping them think about long-term impact, there is a chance to invest now in the technology and organizational infrastructure nonprofits will need for the next decade, not simply the AI tools they might use this year.

The distinction matters because much of what I am seeing right now is AI being layered onto already fragmented systems.

The divide forming around AI in the nonprofit sector is not merely technological. It is structural.

For decades, many nonprofits have been asked to keep overhead low while simultaneously delivering increasingly complicated services, managing sophisticated funding requirements, measuring outcomes, protecting sensitive information, and responding to complex human and community problems.

Technology infrastructure has often been one of the things squeezed in between. Organizations made do.

They added another spreadsheet. Bought a database when funding became available. Kept an old system running longer than anyone wanted. Stored information in several different places. Created workarounds when systems did not communicate. Asked the person who happened to understand the CRM to become the unofficial technology department.

This was often perfectly rational behavior inside the funding environment nonprofits were given.

As a result, many organizations are entering the AI era with fragmented data, disconnected platforms, inconsistent processes, and very little dedicated technology capacity. Many foundations and funding entities face the same problem as service providers.

AI did not create this problem. It is exposing it.

And that gives donors and philanthropy an unusual opportunity to do something about it.

I am seeing a lot of AI being added to systems nonprofits already use.

Document platforms are adding AI search. Board tools can answer questions about policies and meeting materials. Enterprise automation systems are increasingly being adapted for nonprofit organizations. Much of this can be useful. If an organization already has strong document management, clean information and sensible processes, an AI layer can make those systems dramatically easier to use.

But AI sitting on top of fragmented infrastructure does not make the infrastructure coherent. It makes the fragmentation easier to navigate.

An organization may be able to ask a chatbot when a board member’s term expires. At the same time, its donor data, program information, financial systems, evaluation data, and organizational knowledge remain entirely disconnected. This is helpful; it is not transformational. We’re at a moment when meaningful transformation is possible for many nonprofit organizations.

The more important question is what becomes possible when an organization can connect what it knows, understand what is happening across its work, and reduce the enormous amount of human effort currently spent moving information from one place to another.

The opportunity is to free up headspace and time within organizations so humans can spend more time serving humans.

To seize the opportunity today, we need to ask systems architecture questions from within the nonprofit ecosystem, part of the larger philanthropic ecosystem.

A design problem is also emerging. The nonprofit sector is overwhelmingly made up of smaller organizations. According to the National Council of Nonprofits, 97% of charitable nonprofits operate on less than $5 million annually, 92% on less than $1 million, and 88% spend less than $500,000 a year. Yet much of the sophisticated AI and automation infrastructure being developed today starts with an enterprise operating model and is then adapted downward.

We are effectively designing AI systems around 3%  of the sector and adapting them for the other 97%. Enterprise systems work reasonably well for large nonprofits with specialized departments, technology staff, and the capacity to implement complicated systems. Designing operational infrastructure works less well for most organizations that operate differently within the nonprofit sector.

In a smaller nonprofit, the executive director may also be the lead fundraiser, partnership manager, and keeper of institutional history. One person may handle development, communications, and operations. Program staff may serve people while also maintaining the data required by several funders.

Most organizations in the nonprofit sector do not need miniature versions of enterprise technology. They need technology architected around the reality of smaller organizations. Often that means fewer systems, better connections between them, less administrative burden, and access to technical expertise without hiring a full technology department.

Without that investment, I worry that AI will widen an existing divide. The best-resourced organizations will use it to rethink how they operate. Everyone else will accumulate inexpensive tools that make individual tasks faster while the underlying organization remains fragmented.

This is happening at a time when many organizations have very little financial room for major infrastructure investment. Recent research highlighted by Candid found that 39% of nonprofit leaders reported operating at a deficit in 2025, up from 22 percent in 2022. Expecting individual organizations to finance major technology transformation from already-constrained operating budgets independently is not a serious sector-wide strategy.

The infrastructure problem is solvable. And we are at a moment when the tools, talent, and philanthropic capital exist to do something meaningful about it.

Most conversations about nonprofit AI are still focused on productivity.

-How can we write this faster?

-How can we summarize this meeting?

-How can we draft this grant report?

But they are the beginning of the opportunity, not the end. Those are worthwhile uses, particularly in organizations where staff is carrying too much work. But it’s not where the technology is capable of going. I would add that it’s not where the technology is going with each passing day.

The more interesting question is what becomes possible when an organization can use its information to understand patterns earlier, connect knowledge that currently lives in separate places, and adapt while the work is still happening.

Imagine a human services organization that can see changes in demand before a crisis peaks, or an education organization that can bring together what students, families, teachers, and program data are telling it. A community organization might finally be able to see patterns across years of listening sessions, service data, and neighborhood conditions that no staff person has enough hours in the day to hold together.

That is where AI becomes much more interesting. It also becomes more ethically consequential. The goal should not be to collect more information about people simply because technology makes it possible. It should be to build enough responsible infrastructure to create public benefit. The work should preserve agency for the people being served, the people doing the serving, and the people funding it. Better organizational intelligence should not require nonprofits to become more extractive.

For philanthropic advisors, financial advisors, estate attorneys, and others helping donors think about their giving, this creates a practical conversation. A donor does not have to understand AI to fund what makes responsible AI possible. If a donor does understand AI, terrific; partner with nonprofits to create what is possible and fund it.

A better starting point might be to ask nonprofit staff about their daily responsibilities, where operational bottlenecks exist in serving the people they serve (this is why nonprofit organizations exist), and what would significantly improve their service delivery model.

For one organization, that might mean modernizing technology patched together over many years. It might mean cleaning and migrating data so information is usable. It might mean connecting systems that currently require staff to move information manually, strengthening cybersecurity, or bringing in technical leadership the organization could never afford to hire full-time.

The important thing is not the particular investment. It is the willingness to fund infrastructure as organizational capacity. Funding needs to recognize that “the work” is complex adaptive.

Individual donors may have a particularly interesting opportunity here. Institutional philanthropy can move significant resources, support cohorts of organizations, and help establish shared infrastructure across a field or place. But an individual donor can sometimes do something institutions find surprisingly difficult. They can decide.

They can make a significant multi-year investment in an organization they know well without creating a new initiative, application process, or reporting structure around it. For a smaller nonprofit, one committed donor could make an enormous difference.

The donor does not need to fund “AI.” In many cases, the most valuable AI investment may be fixing everything underneath it. There may be no ribbon cutting. No program named after anyone. No photograph of the database migration. Good. Be boring.

Two years from now, staff may spend less time hunting for information. Leadership may see problems earlier. Programs may respond more quickly. Institutional knowledge may survive staff transitions. Boards may govern from better information. Technology may remove administrative burden instead of adding to it. This is the kind of durable impact many funders say they are seeking.

There is a version of the next five years where nonprofits accumulate AI the same way many accumulated software. A fundraising tool here. An AI assistant inside document storage there. Another platform for the board. Another for communications. Another for programs. Each one is useful. None of them are solving the underlying problem. We already know where that path leads.

A better opportunity lies ahead. AI is giving philanthropy a reason to revisit a decades-old infrastructure deficit at exactly the moment when doing so could change what nonprofit organizations can do for a long time. That feels worth funding.

Not because every nonprofit needs sophisticated artificial intelligence.

Not because every organization needs more technology.

And certainly not because AI is fashionable.

Nonprofit operational capacity work is worth funding because nonprofit staff have spent years compensating for weak infrastructure with human effort. They remember what the systems do not. They carry information between departments. They recreate processes from memory. They search through folders and emails. They keep aging systems running because replacing them feels impossible and is rarely funded.

AI can make some of that type of work easier. But the larger opportunity is to build organizations where less of that compensation is necessary in the first place. Organizations where information moves more coherently, institutional knowledge is easier to retain, leaders can see more of what is happening, and people have more time for the parts of nonprofit work that should remain profoundly human.

If donors invest in this space now, the benefit will extend well beyond whatever AI tool happens to be popular this year.

Shanon Solava is a certified Independent Philanthropic Advisor. Her practice focuses on the human dimensions of wealth stewardship, philanthropic strategy, and values-centered governance for individuals, families, foundations, and philanthropic institutions. Reach her at [email protected]

Photo by Sacre Bleu on Unsplash