AI Cohort Program 2.0 | Key Learnings & Insights

Context

Building on the lessons from the AI Cohort 1.0 pilot, AI Cohort 2.0 focused on supporting NGOs in moving beyond AI exploration towards building, testing, and deploying solutions to real-world challenges. The cohort brought together seven organisations working across education, civic engagement, and NGO Platforms, providing implementation grants, technical mentorship, capacity building, and responsible AI support. Over four months, participating NGOs developed and piloted AI applications integrated into their existing programs, strengthening both organisational capability and the evidence base for AI adoption in the social sector.

Program objectives

• Support NGOs in designing and piloting AI solutions aligned to program needs.
• Provide hands-on mentoring and structured guidance.
• Embed Responsible AI principles early in the design process.
• Generate patterns and insights to influence Tech4Dev’s AI platform, Kaapi.

Program design

• Duration: Four months (March – June 2026)
• Participants: 7 NGOs selected from 42 applications, representing education, health, youth development, and civic engagement.
• Activities: Virtual learning sessions, one-on-one mentoring, technical deep dives, responsible AI sessions, and two in-person convenings.
• Mentorship: Each NGO is paired with technical and domain experts to support strategic thinking and solution development.
• Responsible AI Support: Partnered with Tattle to build awareness of AI risks, bias, and responsible practices.

Key elements in the AI cohort

• Four NGOs (Madhi Foundation, U&I Trust, Inqui Lab Foundation and Samanvay Foundation) received funding to hire technical expertise or dedicate internal capacity to develop and pilot their AI solution.
• Building on the monitoring approach established in AI Cohort 1.0, we continued to use a mentor report card in Cohort 2.0. Each month, mentors assessed every participating NGO across five defined parameters.
• Some mentors conducted in-person visits to participating NGOs to understand organisational workflows better, engage with users, and validate use cases.
• Based on lessons from AI Cohort 1.0, we introduced mid-cohort mentor reflection sessions to gather feedback. We also collected mid-cohort feedback from NGOs to identify areas for improvement and make adjustments.

NGO use cases

The second cohort built on the momentum of the first, with organisations moving beyond identifying opportunities to developing and testing AI solutions within their programmes.

• Madhi Foundation focused on supporting teachers and instructional coaches with timely, curriculum-aligned guidance. Their AI-powered lesson planning assistant provides contextual recommendations tailored to state curricula, helping educators access reliable resources and make better instructional decisions. Following development and internal testing, the solution is now being piloted with teachers in one district to gather feedback and refine the experience. 
• U&I Trust addressed the challenge of supporting volunteers in its teaching programme. They developed a WhatsApp-based AI Coach that brings together lesson preparation, attendance, and class support into a single conversational workflow. The first pilot with volunteers validated the approach’s usefulness, and the team is now expanding the chatbot to provide more personalised coaching based on students’ learning needs. 
• Reap Benefit explored how AI could support young changemakers when mentors are not immediately available. Drawing on thousands of past citizen action records, they built an AI coach that helps Solve Ninjas think through civic and environmental challenges, suggests practical next steps, and escalates conversations to human mentors when required. The team is now improving response quality through evaluation systems and curated knowledge datasets. 
• Avanti Fellows continued to refine the personalised feedback system they developed during the previous cohort. This phase focused on improving the accuracy of AI-generated feedback, reducing grading errors, and streamlining mentor workflows. Alongside improving the underlying AI models, the team integrated the solution into their learning management system, creating a more seamless experience for mentors and students. 
• Samanvay Foundation continued its work on simplifying technology adoption for NGOs. Their AI-assisted configurator helps organisations translate programme requirements into configured Avni applications, reducing the time and technical expertise needed to set up digital data collection systems. The team has successfully automated standard configurations and is now extending the solution to support more complex programme designs. 
• Inqui-Lab Foundation expanded its AI-powered student assessment platform to provide structured feedback on innovation projects. The solution evaluates student submissions against defined innovation parameters while keeping teachers involved in validating outputs through a human-in-the-loop process. After validating the system against benchmark datasets, the team is now testing it at scale to assess its performance and the quality of feedback. 
• Glific developed an AI assistant that helps NGOs navigate the platform effectively by answering questions, providing contextual guidance, assisting with flow creation, and escalating queries to support teams when needed. The solution is currently being piloted with partner organisations, with ongoing refinements based on user feedback and automated evaluation of response quality. 

Across these use cases, the common thread was moving from experimentation to implementation, building AI solutions that fit into existing workflows, reducing manual effort, and ensuring that people remain core to decision-making while AI supports them.

Key Learnings

1. Build AI around existing programmes – The successful projects addressed challenges within existing programmes rather than creating new initiatives. This made adoption easier, strengthened organisational ownership, and increased the likelihood of sustaining the solution beyond the cohort.
2. Technical bandwidth determines progress – Technical bandwidth was a key factor in determining the pace of progress. NGOs with dedicated technical resources were able to build and iterate faster, while others required more time and hands-on support. Future cohorts will tailor technical assistance based on each organisation’s capacity and readiness.
3. Iterate early and often – Teams that tested prototypes with users and refined their solutions based on feedback made faster progress and built more practical applications. Experimentation helped validate assumptions and improve usability.
4. Sustainability requires continued support: the cohort enabled NGOs to develop and validate AI pilots, but moving from prototype to long-term implementation requires ongoing technical, strategic, and organisational support. 

Highlights

1. Presentation preparation through dry runs

One of the biggest improvements from the previous cohort was the structured presentation support. Each NGO participated in two presentation dry runs before Demo Day, allowing them to refine their narrative based on the intended audience, simplify technical explanations, and stay within the allocated time. This resulted in stronger, more engaging presentations that communicated both the problem and the impact of the AI solution more effectively.

2. Faster Experimentation and Iteration

Several NGOs adopted a “build, test, learn” approach instead of waiting for perfect solutions. By testing prototypes early with end users and incorporating feedback into subsequent iterations, teams made meaningful progress faster.

Misses

a. Effective use of Grant Money and Resources

Four NGOs (Madhi Foundation, U&I Trust, Inqui Lab Foundation and Samanvay Foundation) received a grant to hire an AI engineer, engage a consultant, or dedicate internal technical capacity to the project. While some organisations were able to manage with their internal tech capacity, others took over six weeks to onboard technical resources, reducing the time available for implementation during the cohort. Despite the funding, many NGOs continued to face technical bandwidth constraints, and we had limited visibility into whether the grant was ultimately used to build dedicated technical capacity.

b. Mentor feedback

• Set clearer expectations up front: Align mentors and NGOs on roles, responsibilities, expected time commitments, and delivery milestones before the cohort begins.
• Strengthen use case definition: Spend more time at the start of the programme refining problem statements and implementation plans to reduce ambiguity during execution.
• Increase early engagement: More interactions between mentors and NGOs in the initial weeks would help establish momentum and build stronger working relationships.
• Provide implementation support: Some NGOs expected greater technical and product development support from the Project Tech4Dev team and from the mentors, particularly for execution.
• Extend post-cohort engagement: Creating an alumni community and holding space for check-ins would enable organisations to share lessons learned, track implementation, and continue supporting one another beyond the programme.

Moving Forward

The second AI Cohort reinforced the value of combining technical expertise, mentorship, and peer learning to help NGOs adopt AI in meaningful ways. Building on these learnings, future cohorts will place greater emphasis on early use case validation, tailored technical support, faster onboarding of technical resources, and continued engagement beyond the programme. We also plan to strengthen the alumni network, encourage knowledge sharing across cohorts, and continue developing reusable AI solutions that can benefit the wider social sector. By refining the cohort model and building on the momentum created, we aim to help more organisations move from AI pilots to sustainable, scalable implementation.

Blog posts for reference:

• From 48-Hour Bottlenecks to 500 DIY Recipes: Our Reflections on Building an AI Civic Mentor | by Gauthamraj Elango | Jun, 2026
• AI Cohort Program 2.0: Empowering NGOs to Scale Impact with AI
• AI Cohort:  Figuring it out together as we go
• Kicking Off AI Cohort 2.0
• AI Cohort Reflections: Design ideas & questions
• Reflections from the AI Cohort 2.0

• T4D AI Cohort 2: Resuming the Avni AI journey
• We Went to Rishikesh to Build an AI Coach. We Came Back With Something Better.
• Avanti Fellows | Experience in AI Cohort 1.0
• Madhi Foundation | Kick-off workshop – AI Cohort 2.0
• Avanti Fellows – AI Cohort 2.0 Reflections