WAY OF THE FUTURE
Nonprofits are embracing AI. Are their funders ready?
Within eight months, an AI-powered chatbot helped reservist families access $31 million in government benefits; funding it took two years
Michael Giladi/Flash90
Illustrative. Israeli reserve soldiers meet with family members during an appreciation event on an army base in northern Israel, on Jan. 29, 2024.
In eight months, a WhatsApp chatbot called Raboti found $31 million in government benefits that Israeli reservists and their families are entitled to, and which most of them hadn’t claimed. Persuading institutional funders to support the chatbot took two years.
Raboti, built by the Israeli nonprofit PollyLabs, asks a few basic questions, works out what a family qualifies for and then, as PollyLabs’ chief partnerships officer Alina Shkolnikov Shvartsman puts it, “it keeps nudging you and nudging you endlessly.” It tells families which document is missing, shows them what it looks like, and checks back the next day to ask whether they’ve uploaded it. For families worn down by bureaucracy, that persistence is the point.
For one user of the chatbot, a reminder to file for a benefit they had discussed days earlier felt like more than a notification. “It felt like someone was actually seeing me and checking in on me,” the user reported. Since its launch in March, Raboti has reached nearly 7,800 people and helped them claim more than $1.3 million from the state’s reservist assistance fund and Israel’s National Insurance Institute.
The people building tools like Raboti say the technology is the easy part. The harder part is the funding system around them, and many donors haven’t caught up.
The data suggests their experience is common. A 2025 survey of U.S. foundation and nonprofit leaders by the Center for Effective Philanthropy found that only 11% of foundations provide any support for AI implementation, financial or otherwise. Fewer than 1 in 5 nonprofit leaders said a funder had even raised the subject with them. And in Fast Forward’s 2025 AI for Humanity Report, 84% of nonprofits already using AI said more funding is what they need most to keep that work going.
“The mental, financial and budgetary flexibility of donors needs to catch up,” Shkolnikov Shvartsman told eJewishPhilanthropy.
Raboti’s own path shows why. For its first two years, its research and development were paid for almost entirely by donors who made their money in tech and venture capital. To them, a chatbot for reservist families was obvious. Shkolnikov Shvartsman describes them as a new generation of philanthropists: very comfortable with AI, but not yet sophisticated philanthropists. Traditional donors asked different questions. How would it work? Why WhatsApp? How could a bot stand in for a case manager? The social sector, she said, is used to helping people by sitting them down with a case worker to answer a long list of questions. Raboti’s alternative is 10 questions by text message, at any hour of the day.
The institutional money came later. UJA-Federation of New York made a one-year grant, beginning in July, to fund Raboti’s final phase of national scale in Israel. “Israel’s reservists and their families have made enormous sacrifices,” said Itzik Shmuli, UJA-Federation’s director general in Israel, calling Raboti an “innovative approach to connecting families with the financial assistance and services they need.” The goal is to reach 100,000 users, about 10,000 of whom PollyLabs hopes will complete an application, unlocking roughly $5 million, or about $500 per family.
The Natan Fund’s giving circle had admired PollyLabs’ work for some time, “but hadn’t yet found the right fit within our grant parameters,” said Avery Robinson, a Natan Fund program officer. That changed with a proposal developed with the IDF Reservists’ Wives Forum, which Robinson called “not only well aligned but inspiring.” It was, he said, what Natan looks for: “an early-stage, creative intervention with the potential to make an outsized difference quickly. Here, that meant an AI tool, delivered through WhatsApp, that helps reservist families navigate the benefits they’re entitled to.” Many members, he added, “chose to increase their contributions because they saw how far this grant could go.”
Even for funders ready to say yes, there’s a catch. Conventional grants assume a plan will hold for a year, but AI tools rarely hold still that long. A typical grant application will ask for budget line items up front and for a year or more. With AI, that plan can be out of date within months. “A system could come to market in two months at half the price,” Shkolnikov Shvartsman said, “or at twice the price but 10 times more effective.” When one application asked for a line-by-line budget naming each system, she said she couldn’t honestly say whether a line would cost $1,000, $2,000, or $10,000.
Raboti’s own costs show how quickly things move. Its cost per user has fallen from about 50 cents to about 20 cents as PollyLabs switched to cheaper, better systems. Its first version ran through Twilio, a third-party service, which meant paying a second company for every message sent over WhatsApp. Now PollyLabs connects to WhatsApp directly. A one-year budget that locks in specific tools, Shkolnikov Shvartsman argues, all but guarantees overpaying.
The big technology companies’ foundations have started running open calls for AI projects. “I think it’s great that they understand the potential of AI to be used as a force for good and empower NGOs to lead the charge on that in their fields of expertise,” Shkolnikov Shvartsman said. But the calls draw floods of applicants, and much of the money goes to large projects in the areas that interest the funder. “In all honesty, if I were a small Jewish nonprofit, which we are, I’m not sure it’s worth the investment,” she said. Those foundations also often provide not just funding but their own tools, which can tie a grantee to their systems for the long term.
Rafi Musher has taken the opposite approach: small grants, aimed at specific projects, that can move quickly. An entrepreneur who co-founded iTrek and sits on the advisory boards of the Israeli Policy Fellowship and the Milken Innovation Center, Musher’s goal is to bring hands-on business skills to nonprofits. Instead of folding technology into restricted grants, he has backed targeted AI projects: a tool iTrek built to share trip information with leaders and potential participants; a handbook for IPF fellows; and an update of the Milken Innovation Center’s impact analysis. The goal in each, he said, was to improve the organization’s impact without adding ongoing operating costs.
“Granting additional funds specifically for these projects was the easiest part – the upside is too great,” he told eJP. The harder part was finding “talented individuals willing to take the chance and lead the work.” Small projects moved “much faster than typical budgeting cycles.” As an investor, Musher said, he wouldn’t back a company that wasn’t using AI, “so why should we not do it for our nonprofits?”
General operating support offers another way around the line-item problem. The One8 Foundation in Boston, which describes itself as a hands-on, data-driven partner, gives its grantees unrestricted funding and leaves the details of AI experiments to them. “I don’t go to the level of an experiment they’re running with AI,” Linor Zeilig, a portfolio manager at One8, told eJP. Its message to grantees is “to experiment in a thoughtful, responsible way, but definitely to go ahead and not wait,” starting with small pilots that can adapt, “because technology that seems cutting edge today can be irrelevant in six months.”
Grants like Musher’s can also pay for something traditional funders rarely support: the work that comes before a finished product. Dana Ginzburg, executive director of the Israeli Policy Fellowship, which works with Israeli graduate students at American universities, led a handbook project Musher funded. Before anything was written, her team used AI to sift through four years of the fellowship’s records, from past activities and documented outcomes to correspondence, surfacing which approaches had worked and why. Then IPF alumni tested those findings against their own experience and turned them into practical guidance for new fellows. “AI could have produced a polished summary of four years of activity, but a summary alone would not have met that need,” Ginzburg said.
That early work is easy to underfund because it doesn’t produce a deliverable. In her earlier consulting work, Ginzburg recalls a client that asked for a dashboard to measure its grantees’ impact. Weeks in, the team hadn’t built a single chart. It was still working out what “impact” meant, what could realistically be measured and where the data lived. The dashboard came eventually, and it worked because of those weeks of preparation. “AI can make building a tool much faster,” Ginzburg said. “It doesn’t remove the need to work out what that tool should help you do.”
Much of this work also lands in the back office, the part of a nonprofit donors least like to pay for. Bar Pereg, PollyLabs’ founder and CEO, was frustrated by how much his organization paid outside firms to handle U.S. state charitable registrations — a slow, opaque process required of most U.S. nonprofits that fundraise across states. PollyLabs’ answer, PollyReg, launches next month: An organization enters its information once, and the tool fills in its state filings automatically from its IRS Form 990. Five organizations have signed on to try it, and PollyLabs says it will cost about half what outside firms charge. “If you don’t know AI, you wouldn’t know we’re using AI,” Matan Gazit, an AI enablement engineer at PollyLabs who builds its internal tools, told EJP.
What these nonprofits need from funders isn’t glamorous, which may be the problem. Funders tend to want their names attached to projects that look important and seem needle-moving. But these nonprofits are seeking funding for work that is mostly back-office: money for planning before any tool is built, budgets flexible enough to switch systems mid-grant, support for technical talent and small grants that can move quickly.
But the proof is in the payoff, and Raboti proves the potential for massive impact. It required no new legislation and no new government budget. The money already existed; families just weren’t claiming it. PollyLabs aims to unlock $5 million more for army reservists over the coming year. Its next stop is New York, where it is exploring a version that helps families in shelters claim cash assistance on the way to housing vouchers.
“These organizations have so much important work to do,” Musher said. “As funders, we just need to help them stay on it by encouraging the resourcing to try new things, to take the risk on improvement.”