When we launched the Agency Fund Accelerator in 2025, in partnership with OpenAI and the Center for Global Development, we set out to answer a big question: Can generative AI expand human welfare, in the places that need it most?
While we haven’t quite answered that question yet, we’ve made progress. Our inaugural cohort of nonprofit organizations built AI products that serve advice and support to pregnant women and new parents, daycare providers, teachers and students, and smallholder farmers. They also invested in continuous evaluations, enabling them to rapidly iterate and improve their operations. Through this process, the organizations demonstrated that AI can be delivered reliably and with safeguards. The cohort has also begun analyzing whether AI expands user agency and helps people achieve their life goals (or whether it instead strips people of self-authorship).
Above all, the 2025 accelerator validated our belief that local innovators, working alongside government systems, are best positioned to build and harness AI for social impact. The experience also clarified the path forward. We saw that while AI’s potential is universal, the infrastructure needed to scale it safely is complex. The next frontier isn’t just about proving that chatbots help bridge digital divides; it’s about building the robust data pipelines, continuous evaluations, and engineering workflows that allow these systems to evolve at the pace of industry, bringing a sense of agency to millions of people.
This is the focus of the 2026 Agency Fund Accelerator.
We’re launching the accelerator’s second cohort in partnership with visionary funders, including Advancing Health Online, Wellspring Philanthropic Fund, Laidir Foundation, Horace W. Goldsmith Foundation, Ezrah Charitable Trust, Twilio Foundation, Cartier Foundation, and Livelihood Impact Fund.
How We Work: Thematic Clusters and Code
Our program aims to help nonprofits use AI to reach a frontier of cost-effective impact and scale. We believe funding is a critical part of the equation – but other inputs are required. To scale innovative AI solutions, organizations need human capital, social capital, and plenty of data.
So our support is both hands-on and operational. We embed talented engineers, behavioral scientists, and product managers directly into grantee teams. Working in two-week sprints, we co-design and help stress-test early-stage AI solutions – ensuring that technical development is always moving toward tangible welfare outcomes.
This year, we have selected organizations across three sectoral themes: Health, Livelihoods, and Citizen Services. These organizations will convene monthly to share learnings, compare notes, and troubleshoot domain-specific challenges. A health team dealing with patient privacy-related data, for example, will solve problems alongside peers facing similar constraints, rather than learning from a generic AI seminar.
Meet the 2026 Cohort
We have selected seven exceptional organizations for this year’s cohort. Moving beyond isolated pilots, these organizations are building the systems that will define the next generation of social service delivery. We expect this cohort to expand in the coming weeks as we finalize partnerships with additional organizations focused on last-mile economic and agricultural support.
Health Cluster: Re-imagining Care Delivery
Intelehealth (India & Kyrgyzstan): Governments are increasingly using provider-to-provider telemedicine to extend primary healthcare to rural communities. While telemedicine connects community health workers with remote doctors, these primary care settings are often overburdened. This increases the risk of diagnostic and medication errors, which are exacerbated by limited information collected remotely. Intelehealth is developing Ayu 2.0, an AI clinical decision support system that integrates seamlessly into existing government primary healthcare workflows. Ayu structures patient history capture, aids clinical reasoning during consultations, and automates clinical documentation. Early evaluations show that by removing the manual administrative burden, Ayu reduces consultation times by 47% while improving diagnostic and treatment planning accuracy by 12% and 53%, respectively.
Living Goods (Kenya & Uganda): Supportive supervision is critical to community health worker (CHW) performance, but in resource-constrained contexts, supervisory capacity is limited by the number of supervisors available, and the fact that they typically have additional duties beyond CHW supervision. Living Goods’ Next Generation Supervisor App (NGSA) addresses these challenges by analyzing CHW performance data in near real-time to generate targeted tasks for field supervisors, enabling more efficient and proactive supervision. This helps improve coverage and care quality without overburdening the supervisors. To further enhance this model, Living Goods proposes to integrate an AI agent that handles medium- and low-priority tasks on supervisors’ behalf, messaging CHWs directly with personalized, actionable feedback and escalating to human supervisors when necessary. This addresses performance gaps while freeing supervisors to focus on the highest-priority cases, resulting in even broader supervision coverage, stronger CHW performance, and ultimately better health outcomes for the communities they serve.
Maisha Meds (Sub-Saharan Africa): Private pharmacies, drug shops, and clinics deliver the majority of health services across Africa as the first point of care for millions – yet vary widely in their staff’s level of clinical training and access to targeted coaching. Maisha Meds is integrating a “smart” clinical documentation and decision support tool into its pharmacy management software, already used by 5,000 provider sites across five countries. Powered by LLMs, the tool will enable providers to capture better patient histories and receive context-sensitive diagnostic prompts, such as catching potential danger signs or contraindications. The goal is to empower these trusted, last-mile providers with greater confidence and discretion to deliver safer, protocol-aligned care to patients who prefer seeking care through the private sector.
Livelihoods Cluster: Future-proofing Incomes
African Management Institute (Pan-African): Micro, small, and medium enterprises (MSMEs) are the engines of growth across African economies, yet many entrepreneurs lack access to locally relevant, best-practice tools and support to navigate the operational, strategic, and personal demands of building a business. The African Management Institute (AMI) bridges this gap through an evidence-based, tech-enabled approach to business development support that drives growth, strengthens business capability, and creates jobs – with scale built into the model. Having worked with over 60,000 entrepreneurs across 39 African countries, AMI equips African business champions with the practical tools, peer support, and confidence needed to build stronger, more resilient businesses. In turn, these businesses create quality jobs – and these quality jobs drive prosperity and dignity across communities.
Kabakoo Academies (Mali & Togo): In West Africa, connecting young learners to social capital is essential for economic mobility, yet traditional vocational training programs often focus narrowly on technical skills – leaving youth entering the workforce without the networks, mindsets, and guidance needed to navigate the personal and social realities of building a business. Kabakoo is deploying an AI-powered upskilling and social capital system on WhatsApp that is now scaling through government partnerships, including a deployment across public schools in Mali and a partnership with the University of Lomé in Togo. By rigorously testing culturally attuned AI-driven mentorship paired with structured peer collaboration, the initiative enables learners to cultivate relationships, access opportunities, and translate skills into economic mobility.
One Acre Fund (East Africa): Smallholder farmers make complex, high-stakes agricultural decisions under conditions of uncertainty – often without access to locally relevant advice or technical support. One Acre Fund is implementing and evaluating Virtual Agronomist, an AI chatbot built in partnership with Innovative Solutions for Decision Agriculture (iSDA). Piloting in Rwanda, they will validate the tool’s adoption, usability, and operational performance through real-world trials, equipping farmers with timely, data-driven guidance that strengthens their decision-making power over planting, input use, and ultimately their livelihoods.
Citizen Services Cluster: Building Systemic Capacity
Adalat AI (Global South): Across the Global South, over 100 million cases sit pending in court systems, with judges spending more time on administrative documentation than on core judicial tasks. Adalat AI deploys AI-powered courtroom infrastructure, including real-time transcription, multilingual translation, and case flow management, which has demonstrated 2–3x productivity gains across 4,000+ courtrooms in India. By eliminating manual clerical bottlenecks, the system frees judges to focus on delivering timely, quality justice – directly reducing backlogs and delays.
Looking Ahead
The work of this cohort will do more than just improve individual programs; it will help define a blueprint for deploying AI in the social sector. Throughout 2026, we will work alongside these teams to test, iterate, and openly share our learnings. We will also build tooling to reduce the frictions facing nonprofit AI developers. We invite you to follow along as we build the infrastructure for a more agentic future.