Funders, Start Your Prompt Engines
AI is a hairpin turn for the nonprofit sector. How philanthropy takes the curve will shape what comes next.
New research confirms what many of us in the sector have felt: philanthropy isn't sure where to steer when it comes to AI in the social sector.
Nonprofits are actively scaling their impact with AI, but funders are still unsure how to evaluate or fuel that work. Alethea Hannemann (Board.Dev) and Chantal Forster (Annenberg Foundation) shared a piece in The Chronicle of Philanthropy, where nonprofits expressed frustration that funders often stall — not from lack of interest, but from a lack of confidence. Project Evident’s Funding the Future report echoes the growing disconnect.
In my work at Fast Forward, I’m a connector between grantmakers and AI-powered nonprofits. We’ve seen both the promise and the paralysis. For this edition, we compiled advice for how funders can confidently engage with AI grantmaking with insights from Nick Cain (Vice President of Strategy & Innovation at the Patrick J. McGovern Foundation), Shannon Farley (Co-founder and Executive Director at Fast Forward), and Aras Jizan (Senior Program Officer at the Gates Foundation).
Fast Forward is also working on a report on how nonprofits are evaluating and building AI. The goal: help funders move from questions to clarity. If you’re a nonprofit using AI, we’d love your input. And if you know one, please pass the survey along.
You don’t have to floor it — but it’s time to shift out of neutral. Here’s how funders can get into gear and steer towards what’s already working.
What’s your most valuable piece of advice for other funders just beginning to evaluate AI-focused nonprofit proposals?
Before you fund your first AI-powered nonprofit, begin with this advice from people already in the driver’s seat.
Nick says it starts with trusting your instincts. “Strong tech strategy usually mirrors strong overall strategy,” he explains. “You don't need to be an expert to spot a strong AI proposal; you just need to stay curious and ask good questions.”
Shannon pays close attention to the process behind the proposal. “Don’t just evaluate the technology, evaluate the practices behind it. The strongest AI proposals are grounded in good data, bias testing, and real community feedback,” she says. “Fund nonprofits that are building responsibly and can show how they’re learning and improving over time.”
For Aras, the focus is on real-world relevance. One use case he looks for is what he calls “navigation technology” — tools that help people make critical decisions in high-stakes moments, like what to do after receiving an eviction notice or losing your job. “The key is asking: Who's going to use this? In what context? And how will we know if it works?” he says. The best proposals, in his view, are targeted interventions, grounded in user research and built to learn as they go, not solutions in search of a problem.
The message is clear: you don’t need to be an AI expert to know what good looks like. Smart proposals solve real problems, center real users, and show their work along the way.
“Narratives about the future of AI are coming primarily out of big tech; we need to ensure they’re also coming out of civil society. Philanthropic funders serve as crucial catalysts by providing flexible capital and de-risking innovative AI projects that can transform the social sector through solutions designed with and for the communities they serve.” - Laura Maher, Chief of Staff; Director of External Engagement, Siegel Family Endowment
What single question has proven most revealing when evaluating a nonprofit’s AI proposal?
Philanthropy doesn’t need to have all the answers when it comes to AI, but it does help to ask the right questions. A quick diagnostic can save you from a bumpy ride.
Nick shared a simple but necessary question: Why? “Why this tool? Why now? Why do you believe your community wants or needs it?” he said. “Challenging yourself to ask ‘why’ — many times if you need to — uncovers the assumptions baked into a proposal. It helps ensure the project is solving a real, clearly defined problem, not just using AI for its own sake.”
Shannon agreed: “Asking why is the right question. Just ask my toddler.” She also keeps a close eye on safety. One question she often asks: How are you building your AI solution to avoid bias and unintended harm? It’s a powerful way to understand whether a team is building responsibly. “You can go a step further by asking if they have an AI policy. If they don’t, I point them to our Nonprofit AI Policy Builder.”
Aras asks: How quickly will you know if your approach is working, and how will you know? “To meaningfully develop and deploy AI in the social sector, we need to bridge the gap between software’s fast feedback loops and impact evaluation’s slower pace,” he said. “Strong teams can point to specific signals they’ll track within weeks, not months. They understand success isn't about perfect deployment but iterative, systematic learning. And they’re thoughtful about testing responsibly, monitoring for unintended consequences and putting guardrails in place to catch issues before they scale.”
For more questions worth asking — and the thinking behind them — check out The Philanthropist’s Guide to Nonprofit AI Investments. It’s a practical starting point for any funder looking to evaluate proposals with more clarity, confidence, and care.
“As a funder focused on empowering people with the skills and tools to succeed in tomorrow’s economy, we support nonprofits that thoughtfully expand access to personalized learning and are focused on teaching young people how to leverage AI for their own future jobs. By investing in these innovations, we can strengthen people's capacity and unlock a wider array of future opportunities for everyone. As AI continues to evolve, philanthropy has a powerful role to play in shaping it for positive social impact.” - Stephanie Lo, Head of Philanthropy, Endless Network
What resources or networks have been most valuable to you in building your own AI funding literacy?
There’s no manual for funding AI in the social sector, but there is a growing community of people learning by doing.
For Nick, the most valuable insights have come from spending time with the people who actually build technology. “Engineers, product managers, data scientists — understanding how tech gets built has given me a clearer sense of what's realistic, what's risky, and what good execution really looks like.”
Shannon takes a similar approach, rooted in proximity to the work. “I have the advantage of working with AI-powered nonprofit leaders every day,” she says. “I find it most productive to sit down with them and ask what’s worked, what’s failed, what it really costs, and what they would do differently. Then, I talk to funders and ask those same questions.” She also makes time to read, from general AI coverage to sector resources and peer perspectives, like those from the Technology Association of Grantmakers, Tristan Harris, and Raffi Krikorian. “It’s worth the time investment.”
To Shannon’s point, it’s important that the community is transparent about how it’s going — funders and nonprofits alike. Jonathan Petts (Upsolve) and Steven Lee (SkillUp Coalition) shared an article on what didn’t work in their early chatbot deployments, and how they’re now evolving their AI strategies to better meet users’ real needs. In their words: “We don’t have all the answers. But we wanted to share our early learnings in case they’re helpful. We’re sharing our honest assessments to open a conversation.”
That’s the kind of leadership that moves the field forward. No one has the full roadmap — but if we share what we’re learning, we’ll all get better at navigating the curves ahead.
Quick Bytes
Other Sector Stories
Fast Forward is launching new AI and Data Bootcamps for early-stage tech nonprofits. Designed to help teams build smarter, safer, and more scalable tools, the virtual trainings will open for applications in June. Come for AI, data, or both.
FutureHouse just opened the lab to everyone. Their new platform features free AI scientist agents that can review literature, uncover what’s been tried before, and help plan experiments with less guesswork.
Google Cloud and the nonprofit Allen Institute for AI (Ai2) are teaming up to expand access to open-source AI models. The partnership will make AI2’s transparency-focused tools easier for developers and nonprofits to use. Last month’s edition with Mozilla dove into the open-source momentum, and this deal throws even more fuel on the fire.
Jacaranda Health’s PROMPTS platform is answering 10,000 maternal health questions a day, by text and with AI. It has supported over 3M mothers in Kenya, triaging urgent cases for human follow-up. The story is part of Humans of AI for Humanity, a joint series from Fast Forward and the Patrick J. McGovern Foundation.
APN Funding News
New Rising Ventures launched its Responsible AI Funding Cycle, offering $50K grants to early-stage nonprofits using AI to serve the public interest. Applications are open through May 22.
LinkedIn announced a $3M fund to support nonprofits around the world that are harnessing AI to drive economic inclusion.
The Schultz Family Foundation committed $3M to AI-powered career tools for overlooked young adults, backing groups like CareerVillage.org and CodePath.
Robin Hood’s AI Poverty Challenge awarded nearly $4M to nine nonprofits using AI to tackle poverty, including Beyond 12 and Unlocked Labs.
Let’s Talk
I am living and breathing AI for humanity these days. If you are too, let’s talk!
The biggest blindspot that funders have, IMO, is to not invest in the underlying technology infrastructure, open data standards development, and LLM governance and ownership. AI for social good, but powered by the same old Silicon Valley tech, is going to have counterproductive outcomes just like the two pocket thinking on the financial side. What we need now is targeted investment into impact tech, along the entire capital spectrum from grants to VC (to wit, Mozilla Ventures). And a sector wide design conversation about this missing infrastructure. Perhaps something FFWD could facilitate?