What four months with the AI Cohort taught me

I have been a part of the AI Cohort 2.0 journey since my second week at Tech4Dev. I joined the project just as the seven NGOs had been selected, and one of my first experiences was attending the opening workshop in Rishikesh. It was my first opportunity to meet the NGO teams, understand the challenges they were trying to solve, and see the excitement around exploring AI for social impact.

Over the next four months, I had a front-row seat to their journey. From weekly mentor calls and working sessions to field visits and countless iterations of problem statements, I watched ideas gradually evolve into tangible AI solutions. More importantly, I learned that building AI in the social sector is rarely about technology alone. It’s about people, priorities, persistence, and the willingness to keep experimenting.

Every NGO started at a different place

One thing became clear very early on: every organisation’s journey looked different.

For almost six weeks after the opening workshop, most NGOs weren’t building AI solutions. They were trying to identify the right problem to solve. Their mentors worked closely with them to narrow down ideas, understand what was actually feasible, and design a solution that would create meaningful impact.

Some NGOs met their mentors once a week, while others needed more frequent discussions. A few mentors even travelled to the field to better understand the realities of the programme before suggesting technical approaches. Looking back, these conversations were just as important as the AI tools themselves.

Progress wasn’t the same for everyone

Some organisations found momentum quickly. They had the right internal team, were able to hire technical support early, and started building their solutions soon after.

Others took longer. Finding the right technical resource wasn’t easy, and many teams were balancing this work alongside programme implementation, reporting, and everything else that comes with running an NGO.

Watching these journeys unfold made me realise that having a good idea is only one part of the equation. The real challenge is finding the time, people, and capacity to consistently work on it.

When AI isn’t the only priority

This was probably my biggest takeaway from the cohort.

Every NGO that joined genuinely wanted to explore how AI could strengthen their work. But AI wasn’t the only thing competing for their attention.

Programme delivery always came first. Teams were managing operations, and handling day-to-day challenges. Naturally, experimentation often had to take a back seat.

Despite this, every NGO made meaningful progress by the end of the programme. They had clearer problem statements, stronger AI strategies, and solutions that had moved beyond ideas into prototypes or testing.

But it also left me wondering what happens next.

Without the structure of weekly mentor calls, regular check-ins, and cohort milestones, will these organisations have the time and capacity to continue building? Will these solutions become part of their everyday work, or will they remain promising prototypes waiting for the next opportunity?

Designing the next cohort

As I look back on the past four months, I don’t think the biggest question is whether NGOs can build AI solutions. They absolutely can. The bigger question is how we can design the next cohort to make those solutions more sustainable.

Some thoughts that stayed with me are:

• Start before the cohort begins. Give NGOs a few weeks to work with their mentors to refine their use case and, if needed, onboard the right technical support before the programme officially starts.
• Rethink technical support. Many organisations lost momentum because they lacked technical expertise at the right time. Providing more hands-on support early on could help them move faster.
• Create clearer milestones. The first few weeks were largely spent refining use cases. More structured checkpoints could help teams transition into building sooner.
• Build momentum throughout the journey. Many NGOs made their biggest strides towards the end of the cohort. The next challenge is figuring out how to create that same momentum from the very beginning.