Data and AI Bootcamps: Our Favorite Takeaways

AI has early indicators of supercharging the impact of tech nonprofits. Builders are experimenting and learning in real time. But AI adoption isn’t without challenges. It can be expensive, requires technical know how, and raises important ethical questions. For many nonprofits, the potential is clear, but the pathway there is not.

[Side note: we know this because we surveyed hundreds of nonprofits on their AI use. Stay tuned for the full report on September 30th, but the TL;DR: nonprofits need funding and guidance to properly leverage AI for impact.]

And guide is just what Fast Forward’s AI & Data Bootcamps were born to do. Four days of experiential learning – fireside chats with seasoned founders and tech experts, in-depth product demos, and practical, foundation-building activities – supported 185 tech nonprofits. They joined us from nearly every corner of the globe to level up their AI knowledge and skills. If you weren’t one of them (or if you want a refresh!), read on for our biggest Bootcamp takeaways.

Data is the unsung hero of tech nonprofit work. It’s easy to see data’s role in impact measurement - Who is being served? Where can our impact be expanded? But data is the overlooked cornerstone of nonprofit operations. Good data practices enhance fundraising, marketing, and internal tracking systems.

But data can do more. It can – and often should – drive what you build for your beneficiary. Quill.org is a great example. An AI-powered nonprofit providing writing support to students, the Quill.org team identified low performance in their grammar APIs. Digging into the data, they learned that students were getting caught up in correcting small grammar and spelling errors. Sometimes that meant not receiving credit for mostly correct answers, frustrating students and teachers alike. Quill.org heard the data loud and clear. Now, they are exploring bundling all grammar feedback together to reduce frustration, increase user retention, and ultimately build a better product. Without this data to tell Quill what users actually want, it would have taken longer to identify the issue and its solution.

AI requires big bets. CareerVillage saw the potential in LLMs. Could AI rapidly scale to support their mission of democratizing access to career counseling? Turns out, yes. With the help of Google.org’s GenAI Accelerator, they got from idea to prototype in just three months and launched Coach, an AI-powered career coaching platform. How did they do it? By shifting the focus of their entire engineering and product team. And it wasn’t just the platform team that had to shift focus. An intensive engineering pivot also requires fundraising efforts and programmatic support to follow suit. But the risk paid off: Coach has now supported an estimated 44,000 learners and supercharged CareerVillage’s work.

But at the same time, big bets require even bigger intentionality. Uprooted Academy supports underrepresented students in accessing college and career pathways. The Uprooted team rightfully worried about inherent bias in off-the-shelf models and how it could negatively impact their users. So, they built their own fixed language model instead. (A fixed language model is a version of an LLM that has gone through an intensive, specialized training process. It gives more stable, and in this case, less biased, outputs than a traditional LLM.) The customization wasn’t without challenges, but it was necessary to meet the needs of their users.

Across the four days of programming, some common themes quickly emerged. Most importantly, how seriously tech nonprofits take their ethical commitments. Responsible AI and data use was interwoven into every session. Tech nonprofits frequently shared how they use community feedback to develop their AI products and data governance strategies. The goal is not just to protect users, but to ultimately serve them better. And, if you’re looking to build an ethical AI policy for your organization, try out our policy builder.

One surprise? How often a “less is more” approach supports data privacy practices. Multiple speakers shared how collecting less data reduces risk. Take Lemontree, a tech nonprofit that makes it easy to access free food. Many of their users belong to vulnerable populations. So, Lemontree doesn’t collect legal names and encourages users to share a one-time location instead of inputting their address. This sensitivity protects Lemontree’s users and increases trust in their product.

Similarly, many speakers cautioned against viewing AI as “magic pixie dust,” stressing the importance of it as a tool, not an end goal. Beyond12’s MyCoachAI extends the organization’s reach, allowing college students to ask for help at any time of day. But human coaches remain in the loop, ready to support any complex needs and support users.

AI and data can supercharge the efforts of tech nonprofits, and we go further when we go together. Shared learning was central to the bootcamps. Speakers provided expertise that only real lived experience creates. Participants asked fantastic questions and connected deeply in breakout rooms. Two founders even laid the groundwork for a project partnership! The tech nonprofit community is such a special place, and we’re grateful for everyone who contributes to it.

A special thank you to all Bootcamp speakers:

• Abhishek Kumar (Cisco)
• Alysia Garmulewicz (Materiom)
• Arghya Bhattacharya (Adalat AI)
• Bryce Bjork (Lenny Learning)
• Corliss Hullett (Salesforce)
• Faisal Lalani (Collective Intelligence Project)
• Jamie Alexandre (Learning Equality)
• Jamie Monville (Quill.org)
• Jared Chung (CareerVillage)
• Natalie Dunn (CareerVillage)
• Renuka Kher (Beyond 12)
• Ryan Jenkins (Compass Ethics)
• Sam Cole (Lemontree)
• Steve Sharer (RipRap Security)
• Valmik Patel (Patrick J. McGovern Foundation)
• Zachary Hanif (Twilio.org)