How The Agency Fund is supporting programs focused on people

When The Agency Fund launched four years ago, it set out to test a provocative premise: that human agency — not technology, not capital, not systems alone — is the most powerful accelerant of development outcomes. Today, the fund is emerging as an influential supporter of locally led organizations that combine behavioral science, technology, and government partnership to scale cost-effective interventions. At the center of its model is a simple hypothesis: When people feel equipped, confident, and in control of their environment, development outcomes improve dramatically. According to The Agency Fund’s website, agency is the “capability to navigate toward the future you want.” And while rooted in social psychology, the fund’s approach is practical — focused on strengthening service delivery systems, enabling rapid experimentation, and embedding technology in government programs that already reach millions. During a Devex Pro Funding briefing last week, Temina Madon, cofounder of The Agency Fund, laid out how her organization’s model works. A new model of philanthropy: The funder-doer Unlike traditional donors, The Agency Fund blends grantmaking with hands-on operational support from a team of engineers, data scientists, product managers, and behavioral researchers. “We like to call ourselves a funder-doer,” Madon said. “We have a team of data scientists and engineers … who embed with our accelerator organizations, and we are working hand in hand, both to understand how to improve the services of each nonprofit, but also to develop public goods. This embedded model grew from frustrations Madon and cofounder Richard Sedlmayr observed in the evidence-based policy movement. After two decades of investing in randomized-control trials and large-scale evaluations, they found that some of the most cost-effective models were coaching, mentoring, and other agency-enhancing interventions that remained underfunded and underscaled. They shifted their approach to focus on organizations that already reach tens or hundreds of thousands of users — often through government channels — to help them rapidly test, improve, and adapt their products. Since its founding, they have brought together a variety of private funds and foundations to create a funding pool administered by The Agency Fund. Members of the funding pool also fund grant recipients directly, supported by The Agency Fund’s proposal and vetting process. Aligned with their emphasis on transparency, complete lists of funders and previous investments is available on The Agency Fund’s website. Rocket Learning: A case study in scale One of the clearest examples is Rocket Learning, an India-based nonprofit working in early childhood education. Rocket Learning now reaches about 350,000 daycare workers and about 5 million children, said Namya Mahajan, Rocket Learning’s cofounder. “One in four government daycare workers in India is now getting Rocket Learning content pretty much every day,” Mahajan said. The platform delivers daily microlearning content via WhatsApp: leveraging a ubiquitous tool while using backend technology to segment users, analyze engagement, and run rapid A/B tests — a method for comparing two versions of something to see which performs best — at scale. The partnership with the Agency Fund has transformed their model. “With the Agency Fund, we … did very deep qualitative work … building the personas of [workers] and figuring out who we were reaching and who we needed different approaches for,” Mahajan said. Perhaps most influential has been the development of Evidential, an A/B testing automation platform cobuilt by Rocket Learning and The Agency Fund. “It’s an automation that allows you to rapidly experiment with variations of your program … and handles the routine statistics and sampling,” Mahajan said. Rocket Learning conducted 15 A/B tests last year alone, and “engagement rates increased by about 15–20%” as a result. As a result of The Agency Fund–Rocket Learning partnership, Evidential is now fully open-source and being integrated into other social-sector products. Building public goods, not just programs A major differentiator of The Agency Fund approach is its commitment to creating open-source public goods — not just improving individual nonprofits, but strengthening the infrastructure underlying social-sector technology. “We really do hope that this is not just something that sits on a shelf, but that we bring it into the fabric of both the software ecosystem for the social sector and the practices of nonprofits by being able to have service providers who can help them,” Madon said. Rocket Learning’s own platform is evolving in this way. “We’re hoping that our platform … can become a public good, and that we can support organizations around the world … We have a pilot going on in Peru … and the hope is that many other organizations can also take this model,” Mahajan said. This approach reflects a shared approach to development: Philanthropy should scale programs, while also contributing to and building ecosystems. How The Agency Fund selects partners The Agency Fund’s sourcing strategy resembles a venture capital funnel. Organizations enter through an open call, with grants around $200,000–$250,000. Those demonstrating strong impact, government reach, and willingness to experiment move into the accelerator. The Agency Fund does not have a specific geographic focus at the current moment, but emphasizes investments in the global south. “We make grants … to new organizations that apply to our open call … but we’re looking for proposals that will teach us something,” Madon said. “We are completely open to working with orgs that haven’t invested much in user research … You just have to be committed to wanting it and to changing culture to make it work.” While early-stage organizations are welcome, Madon said that what matters most to The Agency Fund is the ability to run experiments and learn. They aim to build capabilities — A/B testing, behavioral insights, data pipelines — that persist long after funding ends. Working with governments: The path to scale Both Rocket Learning and The Agency Fund see government partnership as the only sustainable path to population-level scale. “In most countries, government operates at the largest scale … we’re looking for organizations that can achieve that scale by working with government,” Madon said. This philosophy may become increasingly relevant as development financing tightens. With USAID and other bilaterals reducing or delaying funding amid political uncertainty, many NGOs are turning toward domestic government partnerships for long-term viability. For Rocket Learning, this shift has been accelerated by the current funding environment. “This is the first time … we’re feeling that it will be difficult to raise that capital … so we’re deepening the partnership with government,” Mahajan said. Madon sees the same trend across the fund’s portfolio. “Governments in the countries we support are the ultimate decision makers … and that’s probably the way human development will look in future — less foreign aid and more decision-making by local governments,” she said. But engaging the government requires organizations to develop new capabilities such as navigating procurement, aligning evidence with political priorities, and integrating tools into existing systems. This is why the fund’s embedded-team model is so important. As Madon said: “We see ourselves as helping to build foundations … so they can achieve massive scale.” The Agency Fund also takes an unusually open approach to sharing decisions, processes, and expectations. This transparency stems from the founders’ long history in tackling the reproducibility crisis in social science. “We always say that good ideas are built on the shoulders of giants… How can we be as open and transparent as we can so others can build on top?” Madon said. Their partners have taken note. “It really feels like parts of The Agency Fund team become part of your team,” Mahajan said. “The mission truly gets shared.” In a recent collaborative venture, in 2024–2025, The Agency Fund partnered with OpenAI and the Center for Global Development to run an AI for Global Development Accelerator, bringing together organizations with strong evidence bases. “What we wanted to understand is how introduction of AI … might improve outcomes [and] help scale those outcomes,” said Madon. “There were all these common threads … multilingual content, contextualization … and it saves us all years of time.” Cohorts met monthly, Y Combinator-style, presenting progress, breakdowns, and early experimental data. Cross-sector learning was one of the biggest benefits. The next accelerator launches in 2026. Looking ahead: Finding the next generation of ‘lighthouse organizations’ Moving forward, The Agency Fund aims to identify and support what it calls “lighthouse organizations” — locally led institutions demonstrating new models for human development. Rocket Learning is one such lighthouse: a proof point that tech-enabled, agency-driven, government-scale models can be built outside Silicon Valley — and adapted across countries. “We hope to find … 50 lighthouses … that are all shining lights toward the path governments can take,” said Madon. The goal is not to create a portfolio of grantees, but to reshape how development is done: more experimentation, more transparency, more public goods, and — above all — more agency. Want more briefings like this? Let us know — and stay tuned for upcoming live conversations here.