Americans trust nonprofits more than they trust the courts, the government, the news media, or corporations. That trust cuts across party lines, which almost nothing does anymore. Meanwhile, skepticism toward AI keeps climbing.
So what happens when the most trusted institutions in the country start building with a technology a lot of people don’t trust?
I think about Utkarsh Saxena when I try to answer that. Utkarsh runs Adalat AI, a nonprofit that builds AI tools for district courts in India. A lawyer himself, he saw how outdated systems and delays clogged the courts, and he saw how AI could help. He also had to weigh ethical concerns and tech skepticism about AI in the courtroom.
Relying on AI to make judicial decisions was never on the table. Instead, Utkarsh’s strategy was to “build painkillers before multivitamins.” Judges handwrote the record of every proceeding, which drastically slowed the work and left them frustrated and sore. Plus, standard transcription tools missed the regional dialects and legal jargon of Indian courtrooms. Adalat’s solution: an AI stenographer trained on local context.
It was a narrow problem, but one only a purpose-built solution could actually solve. That first useful tool earned the credibility for everything Adalat built afterward. Today, they offer a full end-to-end justice tech stack across over 20% of India’s courtrooms.
“Trust is won and lost person-to-person. Always think of trust in these personal terms, no matter what scale you’re working at.”
- Jimmy Wales, The Seven Rules of Trust
You hear me anecdotally talk about builders like Utkarsh all the time in this newsletter. Our 2026 AI for Humanity Report adds the data behind their stories. We surveyed 119 AI-powered nonprofit leaders across 20 countries and interviewed 20 nonprofit leaders and grantmakers. AI-powered nonprofits, or APNs, are organizations that use AI to deliver their programs and services, making it central to their impact.
At Fast Forward, we’ve helped accelerate more than 120 tech nonprofits, many of them AI-powered. I know this group pretty well. Some of what we found still surprised me.
First, 70% of surveyed APNs weren’t built around AI. They started as regular nonprofits and adopted AI later. They look a lot more like the rest of the sector than like a startup class.
Second, and this one should impact how funders think about grantees: 69% of surveyed AI-powered nonprofits say they served more people last year. Of those, just 13% credit AI alone. Nearly a quarter say AI played a minimal role or none at all. One might assume AI automatically means more reach. Many of these builders judge success not merely by reach, but by what a person walks away with. One leader told us they cap how long students can spend on their education platform, so kids learn what they need and log off. Try imagining that pitch at a social media company! Where APNs report the clearest gains is depth: 55% say AI made personalized services at scale possible.
What didn’t surprise me was the gap between ambition and resources. 90% of surveyed APNs have a plan to scale. Only 24% have the resources to do it. Philanthropy is good at funding the pilot, but now it’s even more crucial to continue supporting what comes after.
If you subscribe to this newsletter, you’re likely among a group of people who want to see that gap get smaller. Grantmakers: my whitepaper co-written with Google.org’s Maggie Johnson and my co-founder Shannon Farley, Realizing the Reset, lays out how to put capital to work. If you’re a nonprofit leader, join us Wednesday, October 7, at 10 a.m. PT for “The State of AI-Powered Nonprofits: A Roadmap for Builders,” a live conversation with leaders from Lemontree, the International Rescue Committee, and Adalat AI. Utkarsh will be there.
“If we don’t shape AI for good, it will be shaped by people who don’t know or care about our problems. If we don’t teach it what matters, someone else will teach it what’s profitable.”
The painkiller framing has a second edge worth naming. Adalat picked a pain the judges felt in their own hands, so the people with the most power to kill the tool became the ones defending it.
It is easy to pick a pain the director or the funder feels instead, usually reporting or grant narratives. Frontline staff then get a mandate rather than relief, and adoption turns into compliance.
Useful test for a first use case: whose week gets visibly easier within a month, and does that person have standing to say so.