When the pandemic halted outreach at Aravind Eye Hospital in India, a Johns Hopkins University team of doctors and technologists stepped in. They built a simple telemedicine platform that let families send eye photos to doctors, who could flag serious problems that might cause blindness if left untreated.
Uptake was quick, and almost immediately the team saw a problem: There were more doctors reviewing images online than treating patients in person. That insight led them to artificial intelligence, and ultimately to the creation of Visilant, a nonprofit that in just a few years developed technology to screen and diagnose more than 30,000 patients in India. The group projects that it will reach more than one million patients within three years.
Visilant is one among a growing number of nonprofits embedding AI into their core programs, according to a new report by the nonprofit accelerator Fast Forward. The study paints a picture of a sector rapidly embracing new tools but struggling to find the philanthropic dollars to keep pace.
“It’s just an incredible transformation that we’ve seen in the sector of how quickly people are thinking about incorporating all these new tools,” said Kevin Barenblat, co-founder of Fast Forward.
The report differentiates between “AI-powered” groups, like Visilant, for which the technology is central to their impact, and “AI-assisted” organizations, which often use off-the-shelf tools such as ChatGPT, Google Gemini, and Anthropic’s Claude for back-office tasks like marketing, grant writing, and workflow automation. More than a third of respondents reported using AI in both ways, showing that once nonprofits get comfortable with the technology, they often expand it throughout their work.
Nearly 200 nonprofits working across issue areas and geographies responded to the survey, conducted this spring with support from Google.org and research led by faculty at Indiana University, University of Minnesota, and Wayne State University. Fast Forward also gets support from companies and philanthropies including the Patrick J. McGovern Foundation, Salesforce, and the Heising-Simons Foundation.
Rapid Growth, Tight Resources
Fast Forward has supported tech nonprofits for a decade, and data from its accelerator program shows just how quickly AI is transforming the field. In 2024, only 13 of 247 applicants identified as AI-powered. By 2025 that number jumped to nearly half, and in 2026 it more than doubled again, to 379 of 782. Today, more than half of Fast Forward’s alumni describe themselves as AI-powered nonprofits.
“This is the way the sector will have to go,” says Shannon Farley, Fast Forward’s cofounder. “In these dynamic times, we all have to do more with less, and AI can unlock efficiencies and impact … There’s a huge opportunity for philanthropists to set the stage for AI that we deserve.”
Even as adoption soars, many groups are still early in their journey — and on tight budgets. Forty percent of AI-powered nonprofits have been using AI for a year or less, and 30 percent have budgets under $500,000, the survey found.
Bigger budgets tend to translate to broader reach. At the smallest scale, AI-powered nonprofits serve a median of just under 2,000 people. Once budgets cross $5 million, they’re serving a median of 7 million people, according to the survey.
The report also found that organizations with more than two years of experience with AI are significantly more likely to open-source or share their tools.
Top Challenges
Funding remains the top barrier. Eighty-four percent of AI-powered nonprofits cite funding for systems, tools, and talent as their greatest need. Nearly half say adopting AI has raised their expenses.
“They’re limited by the vision and comfortability of funders with AI,” Farley said. But she’s starting to see a shift, as grant makers that are “scared of AI” are increasingly open to learning more.
Data privacy and technical expertise are other pressing concerns. Nearly half of respondents cited privacy risks, while 41 percent pointed to the lack of in-house technical expertise.
Nonprofits embedding AI within programs were more likely to have adopted responsible AI practices, according to the survey.
These nonprofits are often working with vulnerable populations, partnering alongside them to design solutions with responsibility, bias, and privacy in mind, said Eunae Yoo, assistant professor at Indiana University’s Kelley School of Business and co-lead researcher on the report.
The report also flagged a looming risk: More than three-quarters of AI-powered nonprofits said they rely on public data-sets to train their models, but many of those U.S. data-sets are at risk of disappearing.
Philanthropy can support nonprofits to help back up and protect their data, Farley said, particularly if they work in a cause area where the data may be removed from public servers.
On-the-Ground Impact
Visilant’s experience shows both the promise and hurdles of scaling up a nonprofit with machine learning. The group built its own AI model using data reflecting the population it serves. A low-cost smartphone adapter allows community health workers to capture clinical-grade eye images. The system analyzes images and patient history to diagnose conditions like cataracts and infections with near-clinical accuracy. From there, the nonprofit provides multilingual counseling and referral instructions. Patients who once traveled hours for exams can now be screened in their own villages, reducing delays, unnecessary referrals, and lost wages.
“Typically philanthropists have wanted to fund a program where we hire someone and we screen a thousand patients,” said Jordan Shuff, a co-founder of Visilant. “Now without the need for additional capital, we can screen a million patients. AI has really unlocked potential for scale and has allowed us to think about our impact in orders of magnitude larger.”
Like many AI-powered nonprofits, Visilant struggled to raise money while developing and validating its model. Funders were reluctant to support unproven technology, but once the group could demonstrate real-world patient impact, fundraising became easier.
Visilant’s revenue grew from $100,000 in 2024 to $2 million so far this year, including a $1.5 million grant from Google.org.
That trajectory mirrors a larger challenge flagged in the report: Many AI nonprofits face a catch-22, needing early capital to prove their ideas but able to attract funders only after results are visible. The report urges donors to break that cycle with earlier, flexible support, not just for programs, but by investing in engineers, data staff, and infrastructure to ensure AI is safe and equitable.
Experimentation Everywhere
About two thirds of AI-powered nonprofits say that chatbots were their entry point into AI, but over time many have adopted more advanced applications like personalizing content or conducting research. The 78 percent of nonprofits from the Fast Forward study that are using AI in an “assisted” capacity, rather than to power it, are more representative of the sector overall, Yoo noted.
Because the voluntary survey focused on nonprofits already using AI in some capacity, it skews toward more tech-forward organizations. Indeed, a recent Chronicle survey of nonprofit approaches to technology found that organizations often struggle to get the money for tech basics, let alone support to adopt more advanced tools.
Still, the trend is clear: Nonprofits of all sizes and budgets are experimenting with AI, testing its limits and potential in real time. How that momentum continues will largely depend on philanthropy, Farley said.
“The onus is on grant makers,” she said. “Nonprofits can only do what funders allow them to do within their budgets.”
