Empowering Nonprofits with AI: Insights and Advice from Our Workshops and Consultations

Last June, we announced that Jen García had joined our organization as AI Education Strategist. The purpose of this role is to harness the transformative power of artificial intelligence in advancing the Foundation’s philanthropic goals, fostering innovation, and supporting the community of grantees of the Foundation and its LA2050 initiative. Read here to learn more about some of Jen’s work with us, and continue reading below for insights and advice, as shared by Jen.

It might seem that if any organizations could benefit from leveraging AI tools to enhance operational efficiency, it could be those in the nonprofit sector. Nonprofit staff are often beleaguered by the myriad of grant applications, each with distinct requirements. Add to that the data analysis, program evaluation, fundraising efforts, donor communications, human resources—the list goes on.

We at the Goldhirsh Foundation set out to discover how to best support our grantees in learning about AI applications. To start, in February 2024, we held an “AI for Good” event, where more than 60 of our grantees convened to learn about AI applications for nonprofits. Feedback from the event highlighted one point above all: grantees wanted more training in AI and a deeper dive into its current capabilities. To empower nonprofits, the Goldhirsh Foundation decided to provide grantees with AI training workshops and to direct AI solution consultations. Over the past several months, these workshops and consultations have revealed several insights about how nonprofits interact with AI.

Insight #1: Nonprofits need to be able to imagine the possibilities before they can ideate solutions for their organizations.

The current capabilities of out-of-the-box generative AI tools in supporting nonprofit staff in their daily tasks have the potential to save them a significant amount of time, but nonprofits typically have limited time to research these tools and their uses. Beyond the products that can be instantly applied, there are tools like OpenAI’s GPT creator, which allows users to build custom versions of ChatGPT based on their own instructions, hand-selected capabilities, and data they upload into its knowledge base. A solution like this has a range of applications useful to nonprofits, but without knowledge of its capabilities, nonprofit leaders can’t imagine how to tailor it to their nonprofit’s needs. This fact was demonstrated in consultations with several of our grantees: those who knew AI’s current capabilities had already thought of applications they’d like help with, whereas those who hadn’t grasped current capabilities couldn’t fathom specific applications.

Actionable advice: Train nonprofits in current AI tools' capabilities to allow for quick identification of opportunities for using resource-saving AI tools.

Insight #2: Customizable AI tools can enhance program delivery.

There are a number of practical uses for customizable tools like OpenAI’s GPT (available to paid users, but they offer a nonprofit discount). Instead of constantly uploading relevant documents to each prompt thread, one can instead upload all relevant documents into their bespoke GPT. One valuable use, imagined by one of our grantees, is a GPT tailored to their organization’s curriculum, pedagogical approach, and the particular needs of the community they serve. Now, instead of scrambling to create a new lesson plan when certain materials are not available, they can prompt the GPT to reimagine the lesson plan without those materials – and even make modifications like designing the lesson plan to be delivered in a park instead of a classroom.

We will continue to assist our grantees to develop AI applications, such as a GPT for data analysis to support impact report writing, program evaluation, and the making of more informed decisions. We’ll also create a GPT to assist in the revision process of outdated institutional policies, and a GPT to answer users’ questions about the programming available in various seasons of their offerings.

Actionable advice: Encourage nonprofits to explore custom AI solutions like GPTs tailored to their mission and data.

Insight #3: Responsible AI adoption is key.

In both our AI workshops and consultations, concerns around data privacy, bias, and responsible AI were top-of-mind. In the workshops, we reviewed AI ethics and responsible AI practices, noting steps nonprofits can take to secure their data— including removing personally identifiable information before using a large language model (LLM)—and other data governance best practices. We also discussed steps nonprofits can take to determine how to methodologically and strategically adopt AI tools, such as designing an AI policy with their board and staff. We surveyed the responsible AI pages of the most popular AI tools, identifying their scorecards in areas like hallucinations, bias, and disallowed content, along with other areas of concern.

Actionable advice: Focus on ethical AI practices, ensuring nonprofits understand privacy considerations and the need for responsible tool implementation.

Insight #4: Reassuring nonprofit leaders that we (humans) are learning AI together sets them at ease.

We have observed that many nonprofits have not attempted to use AI tools due to apprehensions around adopting new technology. More generally, the nonprofit sector is known for being “slow adopters” of tech. Staff shortages, financial instability, and mounting needs of beneficiaries—in addition to tech-adoption anxiety—have kept many nonprofits from utilizing tech. It’s time to change that narrative, especially in light of the capabilities of AI tools and the opportunity to use them to enhance our impact.

And, we humans are adopting this tech together as it evolves. Even the top researchers can’t keep up with the hundreds of research papers published weekly, so, if you’re just learning how to use these AI tools, you’re in great company. Despite common tech anxiety, nonprofit leaders should remember that everyone—from researchers to industry professionals—is learning and adopting AI tools together. We are part of a global community of learners, so be encouraged and share as you learn.

Actionable advice: Remind nonprofits to not worry about falling behind and instead start learning by taking advantage of free AI courses from Google, Microsoft, DeepLearning.ai, DataCamp, and others.

Bonus Insight: Emerging AI agents are set to revolutionize nonprofit solutions.

And…We have a “bonus” insight, which did not arise from workshops and consultations in recent months, but is critical to mention as we look ahead to how AI is evolving.

AI agents—which Sam Altman of OpenAI calls the “third level” in the capabilities of AI in this new Intelligence Age—are in their nascent stages emerging as powerful, programmable tools that could revolutionize how nonprofits develop and implement solutions to address their unique challenges.

“Agents” are software equipped with AI capabilities that can be trained to autonomously make decisions and take actions across various systems to achieve specific goals. Many tech companies are targeting wide adoption of these agents, which would require them to be “no-code” for most users. (Meaning a user wouldn’t need to know any programming language such as Python.) The ability to program AI agents is now widely available through various tech companies, and some, like Salesforce, are heavily investing in them. Google is actively developing agents with “reasoning, planning, and memory” that can “think multiple steps ahead,” and Microsoft announced their integration of agents into Copilot, able to access information across Microsoft apps and storage.

Actionable advice: Adopting current AI tools now and understanding their capabilities will likely lead to a smoother adoption of agents.

We at the Goldhirsh Foundation plan to continue supporting our grantees in the adoption of viable AI solutions, while sharing our learnings with the sector. AI has the potential to provide abundant opportunities for nonprofits to expand their impact and minimize cumbersome processes. As the staff member of one grantee said after learning of an application that will save them hundreds of hours of work, "It's hard to contain my excitement!”