Applications close April 8
AI Impact Scaling Program
Apply today
The conversation about AI in the nonprofit sector has been dominated by two camps. One says the technology is transformative and organizations need to move faster. The other says the risks are real and caution is warranted. Both are probably right. Neither tells you much about what actually happens when a social impact organization tries to scale with AI.
In this article
Since December 2025, we have been working directly with 31 organizations doing exactly that. They come from across the globe - Croatia, Kenya, Malaysia, Peru, Iraq, India, Nigeria, and beyond. They work across health, climate, education, economic opportunity, and humanitarian response. They were selected because they had something proven: a real intervention, real users, measurable results. The question was never whether their work was ready. The question was whether they were ready to scale it with AI.
Three months in, here is what we are seeing - and what is actually helping.
The second cohort of the AI Impact Scaling Program is now open. If you lead a social impact organization with a proven intervention ready to scale with AI — apply here or learn more about the program.
When we asked organizations in our cohort what kinds of support they valued most, the answer was consistent: hands-on working sessions and expert-led guidance tied with 57% each, followed closely by real-world case studies from peer organizations at 52%. General AI education came last.
This tells you something important about where the sector actually is. The appetite for AI literacy content has been enormous over the past two years, and it served a real purpose. But for organizations actively trying to implement, the need has shifted. They are not looking for another introduction to the technology. They are looking for someone who can help them actually build something and work through a specific problem - why an integration is not behaving as expected, how to structure data for a tech partner, how to build internal buy-in for a tool their team is not yet using.
The organizations making progress in our cohort were the ones with access to that kind of specific, applied support. Not generic guidance - direct engagement with their actual implementation challenge.
Pro tip: When evaluating any AI program or support, look past the curriculum. Ask what happens once implementation begins. The answer will tell you more than the program description.
The most common pattern we see is not a technology problem. It is a definition problem. Organizations arrive with strong motivation and a genuine desire to scale with AI. The harder question - what specifically do you want AI to do, for whom, by when, and how will you know it worked - takes longer to answer than most expect.
This is not a criticism. It reflects how genuinely new this territory is for most organizations. The tools have become accessible faster than the frameworks for thinking about them. And in a sector where teams are already stretched, finding time to go deep on problem definition before touching any technology is its own challenge. The problem definition is more than a prerequisite to check off as it will later become the foundation everything else is built on. Organizations that do this work upfront move significantly faster at every stage that follows.
Pro tip: Before approaching any tech partner or program, try to answer three questions in writing: What exactly happens today that AI would change? Who experiences that change directly? How would you measure whether it worked? If the answers feel vague, that is useful information - it shows you where to focus before you start building.
One of the more interesting findings from our cohort was how organizations think about what they bring to a pro bono partnership. When asked what value they could offer a tech partner, 66.7% identified potential for long-term strategic collaboration as their strongest asset - making it the top response by a significant margin. Case studies and testimonials came second at 61.9%.
What this tells us is that organizations understand pro bono as a relationship, not a transaction. And they are right. The partnerships that move fastest are the ones where both sides are genuinely invested - where the organization has a dedicated internal owner who can show up consistently, make decisions, and be a real counterpart to the tech team.
That internal owner does not need to be technical. But they need time - realistically, a minimum of five hours a week during active development - and enough authority to keep things moving without lengthy internal approval cycles. When that person is missing or overstretched, even the strongest tech partner cannot compensate.
Pro tip: Before entering any pro bono partnership, identify your internal owner explicitly. Make sure they have both the time and the mandate. The tech partner will build what you can give them clarity on - and nothing more.
One of the clearest patterns across our cohort is how different each organization's starting point actually is. Some teams are already running sophisticated AI systems and want to go deep on model evaluation and retrieval-augmented generation. Others are mapping their data landscape for the first time. Both are exactly where they should be - they are simply at different points in the same journey.
This reflects something true about the nonprofit sector at large. AI adoption is not happening uniformly - it is happening in waves, at different speeds, across organizations with very different technical histories. The result is a sector where the distance between the most and least advanced organizations is wider than most AI programming accounts for.
This gap is precisely what individual matching addresses. When support is designed around a general curriculum, in a group this diverse, the average might not answer the needs of the most. What actually moves organizations forward is support calibrated to where they are: a tech partner selected for their specific challenge, expert guidance on their specific bottleneck, peer connections with organizations at a similar stage.
The organizations that progress fastest in our programs might not be the most technically advanced ones. They were the ones whose support was most precisely matched to their actual starting point.
Pro tip: If you are evaluating AI programs, ask whether the support is designed around your specific stage and challenge - or around a general curriculum. The former will move you faster, regardless of where you are starting from.
The 31 organizations in our cohort are at different points in figuring all of this out. Some are deep in active development with tech partners. Some are still sharpening their problem definition. All of them are doing the real work - the slower, more organizational, less glamorous work that comes after the AI strategy gets written.
That work does not get talked about much. It probably should.
Yes - you need a validated solution with real users and measurable results. The Scaling Program is designed for organizations that have already proven their intervention works and are ready to multiply its reach with AI.
Jumping to technology before the problem is clearly defined. The clearest predictor of progress in our cohort is not technical sophistication - it is organizational clarity about what the solution needs to do and for whom.
More than most organizations expect at the start. A dedicated point of contact able to commit a minimum of five hours per week is the practical baseline for any meaningful pro bono tech partnership. Below that, progress slows significantly regardless of how capable the tech partner is.
More than they often realize. Long-term strategic collaboration, case studies, social media recognition, and impact reporting all ranked highly when we asked our cohort this question. The organizations that frame pro bono as a genuine partnership - rather than a service they are receiving - tend to attract stronger partners and build more durable collaborations.
Three signals worth checking: you have a proven intervention with measurable impact, you can articulate specifically what AI would change about how it works or reaches people, and you have someone internally who can own the implementation process. If any of those three are missing, that is the place to start.
