AI Cohort Reflections: Design ideas & questions

Running AI cohort 1.0 was the first project that I worked on at Tech4Dev. This week we concluded our second AI cohort with a showcase by the 7 participating NGOs in Bangalore. The objective of the AI cohort program at Tech4dev is – In a world full of conversations about AI, can we provide practical, hands-on support to NGOs? Building with them is the key goal and not asking them to build. We don’t want to hand NGOs a map and wish them luck but try to walk the path with them, one experiment at a time.

The mindset/way of working in the cohort is –

1. Experiment a lot, fail fast, learn and iterate
2. Create a POC, test it with users and bake in their inputs throughout the journey
3. Seek guidance from mentors, what’s new in AI that we can leverage
4. Share with each other, build on top of what others have already built, let’s not re-invent the wheel

The cohort design is –

1. Onboard 7-8 NGOs through an open application process 
2. Map them to mentors who are experts in tech/AI
3. Two in-person meetings and virtual guidance by mentors over a course of 3-4 months
4. An unrestricted grant along with LLM credits shared with NGOs to enable experimentation
5. The cohort becomes a peer group to share and learn from each other

What success looks like –

1. NGOs have seen a spark, where they have started using AI and it feels a bit more familiar
2. NGOs feel confident that they can keep learning, use AI confidently with some guidance
3. If we have been able to identify patterns and realize the sector needs better

The AI cohort program is designed to give NGOs a spring board into the AI rocketship.

What we are still trying to improve on –

Human motivation has always been a conundrum for me. What drives us to do what we do is really difficult to gauge. I try to put this framework to think through while designing cohorts. (It keeps evolving, please let me know if your thoughts/versions of it)

I feel that if we can answer the following questions for each individual, we will be setting them up for success  –

1. Goal setting – Where do want to reach?
2. Role setting – What is my role in getting there? How do I contribute?
3. Motivation – Why should I travel all the way?
4. Resources to reach – Do I have tools and resources?
5. Check-points – Are there small checkpoints to to tell me how far I still need to go? 
6. Guidance – If I take a wrong turn, will someone guide me?
7. Reminders and nudges – Will I have reminders to stay on track?

How this looks like, in the context of the AI cohort program is –

1. Goal setting – Pilot the solution with 50-100 people
2. Role setting – 100% bandwidth of a technical person is needed. They need to be invested in the program. Meetings, weekly updates and sync calls are setup.
3. Motivation – Continued support + graduation to next cohort
4. Resources to reach – An unrestricted grant, LLM credits, mentor support through the program
5. Check-points – Demo presentations by NGOs. Mid-point check-in discussions.
6. Guidance – Mentors provide continuous support to NGOs (2-3 hours per week)
7. Reminders and nudges – Regular check-ins and follow-ups were backed in the design along with NGO report cards filled monthly by mentors

Where I feel we still need to refine the program is –

1. Role setting – Although agreed upon, the truth is that NGOs are really stretched on bandwidth and amidst competing priorities – experimentation and POCs take a back seat before the live fire fighting they need to work on. Can we explore a model where we provide them with tech experts for the cohort and ensure a smooth transition to their internal team through a grant structured to support that? Should we increase the cohort duration as well? How can we better align the AI use cases to the priorities that NGOs are already working on?
2. Resources – How do we ensure that resources are used at the right time for the right purpose of the cohort?
3. Check-points – These are limited to the cohort duration. How do we ensure we have built capacity in the NGO to take the work forward and that the cohort is not a one off intervention? How do we ensure sustainability beyond the cohort?