The Playbook on AI-Powered Nonprofits
Have you met AI?
Of course you have. From the app reminding you to hit your step goal to the platform that predicts your perfect summer playlist, it’s everywhere. Yet for many nonprofits, AI still feels unfamiliar…maybe even daunting. It can be tough to connect the dots between innovative tech and a nonprofit mission rooted in empathy, community, and social justice. How does a tool primarily designed for prediction and automation fit into work centered around people?
But here’s the thing: AI doesn’t have to be at odds with your organization’s values. When used responsibly, it can help you streamline repetitive tasks so you can do more of what you’re already great at — driving social impact.
Whether you’re an AI-powered nonprofit (APN) or a nonprofit exploring AI to improve your internal ops, this guide has your back.
“We believe that the benefits of these new AI technologies cannot remain locked and accessible to only a few. We must work to ensure everyone benefits equally from them.”
- Gemma Turon, Co-founder and CEO of Ersilia
What to Expect From this Playbook
Ready to explore the world of AI through a nonprofit lens? Here’s what’s coming your way:
-
The Basics of AI: A primer on what AI is and why it’s a helpful tool for nonprofits to increase impact and efficiency.
-
The Current AI-Powered Nonprofit Landscape: Dig into how AI-powered nonprofits (APNs) are leveraging AI today, with insights into the different ways organizations are integrating AI into their missions.
-
Getting Started: Key Questions to Ask: Critical factors to consider before adopting AI to ensure it aligns with your nonprofit's goals and values.
-
The Basics of AI Models: A breakdown of AI, why most nonprofits use off-the-shelf models, and how to customize them for your organization’s needs.
-
Lessons from APN Leaders Who Have Been in Your Shoes: Learn from the experiences of successful APN leaders as they share practical insights and best practices for implementing AI.
-
Case Study: Using AI to Bridge Language Gaps for Refugees and Asylum Seekers: A real-world example of how Tarjimly uses AI to translate for refugees by combining AI efficiency with human oversight.
-
How to Pitch and Fund Your Own AI-Powered Nonprofit: Fundraising 101, including tips from CareerVillage's Jared Chung on how to effectively communicate your AI-powered nonprofit's mission to secure funding.
-
How to Use AI Without Building A Product: Operational AI for Nonprofits: Discover how nonprofits can harness AI tools for operational efficiency.
Definition
AI-powered nonprofit (APN)
A nonprofit that develops its own AI-based solutions, using AI as a core component to achieve social impact. 40% of Fast Forward's portfolio are AI-powered nonprofits...with others on the way.
Definition
Nonprofits that use AI
Nonprofits that leverage existing AI-based tools and platforms to enhance their operations, programs, and impact.
The Basics of AI
You already know this, but we’ll share anyway… At its core, AI is the simulation of human intelligence by machines. AI helps machines analyze data, spot patterns, and make decisions — just like people.
But AI isn’t people. We humans are essential to guiding this technology, making sure it aligns with our values. In the nonprofit space, human oversight is even more essential. Nonprofits work with vulnerable communities and, as such, must do their very best to ensure AI acts responsibly and equitably, enhancing (not replacing) the human touch that makes our work meaningful.
Key Terms
-
AI-powered nonprofit (APN)
A nonprofit that develops its own AI-based solutions, using AI as a core component to achieve social impact. Example: Lenny Learning's AI-powered platform, Lenny, tackles the youth mental health crisis by helping teachers assess student needs and generating personalized mental health lesson plans.
-
Artificial Intelligence (AI)
The capability of computers to perform tasks that typically require human intelligence. You’re using AI if Netflix has suggested your new favorite show or you’ve used facial recognition to unlock your phone. Example: TalkingPoints uses AI to translate between schools and families. It combines in-house machine learning with human translators to provide culturally relevant translations.
-
Ethical AI
AI systems designed and deployed with an emphasis on fairness, transparency, privacy, and accountability. Example: With its AI Coach, CareerVillage holds itself to high ethical standards. How? They put the AI technology in front of the communities they serve, are consistently vocal about its limitations, and then listen and let the users guide them.
-
Generative AI (GenAI)
A type of AI that creates content, like text, images, and sounds, based on patterns learned from existing data. This might be what comes to mind when you hear “AI.” Think Google Gemini, ChatGPT, or Dall-E. Example: Khan Academy's Khanmigo is a generative AI tutor and teacher’s assistant. It offers personalized guidance, feedback, and interactive learning.
-
Large Language Model (LLM)
A type of AI model trained on large datasets to understand and generate natural language. LLMs are an advanced application of NLPs that utilize deep learning. This tool is used for a range of tasks, including answering questions, drafting content, and having conversations. Example: Tarjimly's Generative AI is being used to create an accessible and personalized translation workflow for Tarjimly. Alongside this, a data hub is being developed to help train large language models (LLMs) for low-resource languages.
-
Machine Learning (ML)
A subset of AI where computers learn from data to make predictions or decisions. A simple example is how your email filters spam by recognizing patterns from messages you usually mark as junk. Example: Reboot Rx synthesizes hundreds of thousands of research studies with machine learning to identify generic drugs that can treat cancer.
-
Natural Language Processing (NLP)
A sub-field of AI that enables machines to understand, interpret, and generate human language. NLP is used in voice assistants like Google, Siri, and Alexa, which know what you mean when you say, “play La Isla Bonita.” Example: Learning Equality's tool, Kolibri, uses NLP to align curriculum from 200K educational resources to educational requirements. This makes materials accessible offline and searchable in multiple languages.
AI for Humanity
Stay up-to-date on the latest in AI-driven social impact. Subscribe to the AI for Humanity newsletter today!
Pro Tip
Show Up, Plug In
Attend industry conferences, webinars, and networking events related to AI, technology, and social innovation to meet potential collaborators and funders. Consider plugging into networks like Fast Forward, All Tech Is Human, and more.
The Current AI-Powered Nonprofit Landscape
Considering building an APN? You’re in good company. At Fast Forward, we've been closely following the rise of APNs, and the growth has been nothing short of remarkable. To make sense of it all, we developed a working landscape to connect the dots. In the end, four big categories of nonprofit AI use cases emerged: Structuring Data, Advising, Translating, and Platforms. Excitingly, we found that most APNs don’t use AI in just one way. Their solutions tend to consist of a constellation of use cases. Another key finding was that across sectors, the applications of AI aren’t all that different. That means that regardless of your sector, the paths of other APNs might inspire your own way of deploying AI.
To dig into the landscape, check out our Introducing AI-Powered Nonprofits series in partnership with the Stanford Social Innovation Review.
Ethics
Make your responsible AI practices crystal clear. Show your commitment to data privacy, security, and ethical development to build trust and prioritize transparency. Show how ethical guidelines shape your AI projects, ensuring they align with both safety and integrity. Sharing these details invites trust and underscores your dedication to responsible, thoughtful AI deployment.
Looking for guidance on implementing AI in your nonprofit while adhering to ethical principles? Check out our AI Policy.
Getting Started: Key Questions to Ask
Before diving into AI, you must ask yourself key questions to ensure AI works for – and not against – your mission. Skipping this step could lead to unintended consequences. We have a broader list, which we advise funders to ask potential AI prospects about here.
-
What specific problem is your nonprofit trying to address with AI?
-
Who are the beneficiaries of this solution, and do they want an AI solution like this?
-
How will the AI solution create positive change in the lives of your beneficiaries?
-
Do you have the technical expertise and resources to implement and manage the AI solution effectively? If not, do you have the means to acquire them?
-
Is the downside bigger than the upside? Have you explored potential negative impacts and how your nonprofit plans to mitigate them?
The Basics of AI Models
So, you’ve asked the right questions, and you’re ready to dive in. Start with AI models. At its core, every AI-powered product relies on one. The not-so-secret secret? Most nonprofits (and companies) don’t build their own — they use off-the-shelf AI models. You can, too. The real work often comes in fine-tuning the model for your specific use case and developing a delightful product experience for your beneficiaries. When choosing your model, zero in on the essentials: accuracy, cost, ease of use, and, of course, making sure its policies align with your values.
Deciding between using a pre-trained AI model or fine-tuning one for your nonprofit can feel like a big choice. Here’s a quick breakdown:
Off-the-shelf models
Off-the-shelf models like GPT-4o are ready to go. They’re trained on large amounts of data and can handle common tasks — think answering questions, generating content, or automating routine processes. These models are easy to set up, affordable, and great if your needs are broad. For most nonprofits, this is enough to get started without extra fuss.
Fine-tuning
Fine-tuning a model means customizing it with your organization’s specific data — like donor histories, community-specific language, or hyper-local insights. While this can make the model more precise, it takes more time, money, and technical expertise. Unless your nonprofit’s work is highly specialized, fine-tuning often isn’t necessary.
Overview of Popular Models
“We learned that democratizing access to AI tools isn't enough; we must also teach individuals how to use them. For many, Coach was their first interaction with generative AI. We felt a responsibility to ensure it was a positive experience that set users up for success with AI-driven tools.”
Pro Tip
Scale With Support
Participate in programs that help you scale, like Fast Forward's Accelerator, Tech To The Rescue, GitLab Foundation, Robin Hood, data.org, and more!
Lessons from APN Leaders Who Have Been in Your Shoes
Let’s dive into how to navigate implementation decisions with leaders of AI-powered nonprofits. We’ll hear from:
-
Tarjimly: Uses AI-powered translation tools to break down language barriers for refugees and asylum seekers, ensuring access to vital services and resources.
-
ASAP: A hub for U.S. asylum seekers to access work permits and legal aid, focused on empowering asylum seekers and using AI to navigate complex legal processes.
-
Lenny Learning: A nonprofit addressing youth mental health by supporting teachers in assessing student needs and generating personalized mental health lesson plans using AI.
-
CareerVillage: Provides personalized career guidance at scale to underserved communities. Coach uses AI to connect students with career mentors, job guidance, and tools to succeed in their careers and beyond.
-
Lemontree: An APN that addresses food insecurity by connecting families in need with food support services. Lemontree uses AI to ensure that they are referring the best and closest resources to the people who need them.
Which AI Model is right for your nonprofit?
As we explored above, developing your AI-powered product on an off-the-shelf model is likely the move. Like Tarjimly, which experimented with the most powerful models available through Google Cloud Vertex AI, evaluating factors like accuracy and efficiency and using BLEU scores (a metric used to evaluate the quality of machine-generated text, particularly in the context of machine translation) to measure performance. ASAP prioritized security and self-hosting for sensitive data, choosing models like OpenAI’s GPT-4o. Lenny Learning chose GPT-4o based on its favorable unit economics – it enabled them to scale quickly without a lot of additional costs. CareerVillage developed Coach by bringing together a lot of models, which they selected by considering a range of factors like quality, cost, consistency, bias, and language capabilities. They also reviewed external resources like blog posts, tested different models against their use cases extensively, and finally conducted hands-on trials with real users.
Building Your Dream AI Team
The size and composition of AI teams vary among nonprofits – from the one-person-shop to the full dev team. Tarjimly and ASAP upskilled their existing teams rather than hiring new staff. They transitioned key roles like Lead Engineer and Product Manager to focus on AI-related responsibilities. Lenny Learning relied on a very small team to move quickly (we’re talking one-person small!). Lemontree also started with just one person building and managing the AI system, gradually expanding the team over time. CareerVillage employed a broader range of roles, including engineers, product managers, data analysts, and more, to support the development and maintenance of their AI-powered product.
There are lots of ways to slice it! Bringing on an advisor who is strategic and well-versed in AI and nonprofit can be great for developing an AI product strategy. On the other hand, hiring out to dev firms can be great for executing an AI product build. But we typically recommend hiring an in-house technical team. Why? Given how much knowledge of a tech stack is needed to properly integrate AI, hiring folks to fully focus on the product is often best.
Challenges You Might Face When Implementing AI Solutions
Implementing AI comes with unique challenges. Lenny Learning found it relatively easy to create a basic demo but struggled to maintain high-quality performance across all use cases. These challenges are important to anticipate because AI’s effectiveness depends not just on building a prototype but on its ability to consistently deliver results across real-world scenarios. Being prepared for these hurdles helps ensure your AI solution can meet the needs of all users. Lemontree initially had concerns about how well their AI-generated recommendations would align with human intuition. So, they built tools to compare machine-generated results with human referrals to ensure quality. THE TL;DR: Stuff will happen, so stay flexible.
How to Keep Improving Your AI-Powered Product After Launch
The rapidly evolving AI landscape requires continuous iteration and improvement. To stay ahead, keep an eye on the latest AI trends and tools, experiment with new tech, and actively seek user feedback. By remaining adaptable and curious, you’ll ensure your product keeps evolving along with the technology. Tarjimly keeps improving its AI-powered product by leveraging the Google Cloud Vertex AI platform to test new models and collect user-generated fine-tuning data. They plan to implement features like voice integration as LLMs become faster. ASAP is focused on refining how their conversational flows engage members, saving and analyzing conversations to keep optimizing. Lenny Learning regularly reviews inputs and outputs, refines prompts, and quickly adopts new models as they become available. Lemontree keeps fine-tuning its ranking models (a system that organizes items in order of importance), adding new features as their dataset grows. They’re planning to broaden their AI capabilities.
Funding Forward
If you're looking for ways to stay ahead, subscribe to Funding Forward. It's packed with AI funding opportunities, pro-bono technical help, and helpful content. All curated specifically for leaders like you. Join over 2,400 leaders who rely on Funding Forward.
Case Study
Using AI to Bridge Language Gaps for Refugees and Asylum Seekers
We shared a little bit about Tarjimly above. Let’s go deeper. Tarjimly is an APN committed to dismantling language barriers that stand in the way of better lives for refugees and immigrants. Here's how they use AI to push their cause forward and impact the lives of those who need it.
Tarjimly saw AI's potential to transform translation services, but they also recognized the risks — especially in humanitarian settings where even small mistakes could have serious consequences. To safeguard accuracy, they took a hybrid approach: AI speeds up translations, but human translators ensure precision. This balance allows Tarjimly to provide faster, more dependable translations while maintaining the human touch needed in critical refugee support situations.
How Tarjimly Chose the Right AI Models
As we mentioned above, Tarjimly decided which models to use by experimenting with the most powerful options available through the Google Cloud Platform's Vertex AI. They evaluated models like Claude 3.5 Sonnet and Gemini 1.5 Pro, assessing their performance — including BLEU scores (see definition above!) — to ensure they met Tarjimly's high standards for accuracy and efficiency.
Impact Metrics
The results have been impressive. By integrating AI into its translation process, Tarjimly saw a 3X increase in translation speed and an 18% improvement in accuracy compared to human-only translations (we were shocked, too!). More than 90% of translators using the AI tools reported positive outcomes. These AI-driven translations have facilitated thousands of conversations, helping refugees and aid workers overcome language barriers.
Ethical Considerations
For Tarjimly, the ethical use of AI is a priority, particularly given the sensitive nature of refugee data. The organization follows strict safeguarding standards and conducts risk assessments before rolling out AI tools. Data security and translation accuracy were the main concerns, so Tarjimly collaborated with Google’s security experts and implemented HIPAA-compliant protocols to mitigate risks. Extensive testing with their translator community ensured that AI-powered translations met the organization’s quality standards, maintaining a balance between the speed of AI and the critical oversight provided by human translators.
Challenges Faced
One of the main challenges in implementing AI tools was managing project scope. It was important to avoid over-complicating the process and focus on building a Minimum Viable Product (MVP). This allowed Tarjimly to test their AI-powered translation solution early on, refine the approach, and scale the project in a controlled and efficient manner.Tarjimly’s journey shows what’s possible when technology and compassion come together. It’s a testament to how AI can enhance, not replace, the human touch.
According to a report by Google.org, nonprofits leveraging AI for operational tasks can 3X their speed at half the cost, reallocating resources to mission-critical activities. Side note: check out Nonprofit Gemini Prompt Library for more time-saving tools.
Open-Source AI Models for Nonprofits
Open-source models are AI tools that anyone can access, modify, and share. Unlike proprietary software, they’re free and customizable, giving organizations more control over how they use them.
LLAMA (Meta) and Bloom (BigScience) are two popular open-source LLMs. They’re designed to generate and understand text, much like GPT-4o, but without the barriers of cost and licensing restrictions.
How can nonprofits benefit?
-
Cost-Effective: These models are free to use, allowing nonprofits to save on pricy AI services
-
Customizable: You can modify the models to fit your unique needs, whether that’s language, bias reduction, or specific outcomes.
-
Community Support: Access to a global community of developers who continuously improve the models ensures you benefit from ongoing enhancements
Pro Tip
Tap Into Tech
Collaborate with tech companies who can provide credits, technical expertise, resources, or funding.
How to Pitch and Fund Your Own AI-Powered Nonprofit
You will very likely need some financial support to get your AI-powered nonprofit off the ground. Raising funds for an AI-powered nonprofit (APN) presents unique opportunities and challenges. While the transformative potential of AI can be a compelling selling point, funders may have questions about its ethical implications, sustainability, and real-world impact.
When pitching your AI-powered nonprofit, lead with impact — show funders the difference they can make by backing you. Communicate that AI isn’t just some shiny buzzword but the secret weapon that can make their dollars work harder. Funders want to know their support is making a real difference, so draw a direct line between your AI solution and the tangible, people-centered impact it creates. Make it clear: they’re not just investing in AI. They’re investing in a smarter, faster path to positive change.
Case Study
Coach, a Vision for Equitable Career Guidance
Meet Jared Chung, founder of CareerVillage. He took his nonprofit from a career mentorship forum to an AI-driven career coaching platform. He called the new AI-powered product Coach. With Coach, CareerVillage delivers personalized career guidance at scale, targeting underserved communities. Here, Jared breaks down the basics of getting support for your AI-powered nonprofit.
What was your initial vision for Coach?
The vision for Coach was to create a tool that could act as a personal career coach for anyone, regardless of their background, making world-class career guidance accessible to everyone. We wanted to leverage AI in a way that felt personalized, responsive, and scalable — assisting with everything from mock interviews and resume building to career exploration. Beyond these features, Coach had to understand the nuances of career development and adapt to each learner’s unique needs.
How did you communicate that vision to potential funders?
When communicating with funders, we emphasized that Coach wasn't just another AI chatbot — it was developed in collaboration with over twenty organizations like ASU and Year Up to meet the needs of underserved communities. We highlighted our coalition approach, bringing in subject matter experts and engaging in co-design with real learners. We leaned into our bold vision to transform economic outcomes at scale and stressed the moral imperative of a nonprofit leading this work.
How did you go about securing funding?
We started by aligning our mission with funders who were passionate about workforce development, education, and technology's potential to scale impact. Foundations like the Bill & Melinda Gates Foundation, Google.org, and Salesforce resonated with our narrative on how AI could transform career development and address the gap in high-quality career support.
Were there any specific challenges you faced while fundraising?
Yes, a major challenge was explaining how AI could responsibly serve underserved populations. There's skepticism about whether AI can create equitable outcomes, so we had to prove that we were building something that complements, not replaces, human guidance. We provide concrete examples of our intentional, equity-centered development process, including bias testing and in-house monitoring systems.
Another challenge was addressing concerns about sustainability in a fast-moving field. We emphasized our coalition model and how the tool would evolve with input from the communities we serve, highlighting our team's expertise and track record.
What would you recommend to first-time founders trying to overcome those obstacles?
First, establish credibility by clearly articulating "why you, why now." Be transparent about your technology's limitations and how you'll address them through oversight, partnerships, or iterative development. Humility and a commitment to iteration build trust.
Build a strong community around your vision. Collaborate with nonprofits, schools, and workforce boards. Community support strengthens your product and reassures funders of sustainability.
What advice do you have for others seeking investments for nonprofit AI projects?
Our advice to others is to start with a strong social impact narrative. Clearly articulate your unique value proposition and any early indications of impact. Emphasize strategic partnerships and sustainable plans. Funders are looking for projects that consider the ethical and social implications of AI and approach development responsibly.
Be mindful of where funders are in their AI investment journey. Some are ready to invest now; others are still exploring. Use this as an opportunity to lead their learning journeys—keep in touch and share updates.
How to Use AI Without Building A Product: Operational AI for Nonprofits
You don’t need to be a tech genius or have an AI-powered product to get AI working for you. As Shannon Farley, Fast Forward's co-founder, shared at Google.org's Impact Summit, AI can stretch donor dollars further by optimizing operations and freeing up your team to focus on what matters most. AI tools today are plug-and-play, designed to simplify tasks and boost efficiency without sacrificing ethical standards.
Nonprofits can tap into AI without heavy lifting. Many tools are easy to implement and don't require specialized tech expertise. A great example of this is Google’s Nonprofit Prompt Library, designed to streamline repetitive tasks. With the help of AI, you can save time when automating emails, enhancing donor outreach, and handling administrative duties. Here are some ways to integrate AI into your nonprofit:
Writing Support
AI can streamline the writing process (hello, proposals, and grant reports!) by helping with blank-page syndrome, drafting copy (for you to edit), and reformatting content to meet specific criteria — saving valuable time for teams.
Sample prompt: As part of the grant proposal, draft a 350-word answer to the question: "How will funding help the communities we serve?" Pull information from our website [insert URL] to inform your answer, and ensure that your answer aligns with the values of the funder [insert link to funder's website]. Keep the tone formal, but make the language simple and easy to understand with short sentences. Use numbers to back up any claims when needed.
Data analysis
AI-powered tools can process large volumes of data, revealing trends and insights that might otherwise go unnoticed. This can help your nonprofit make smarter, data-driven decisions. For example, AI can quickly sift through program outcomes and surveys to reveal patterns and provide a suggested path forward. Another example: AI can forecast future fundraising trends or donor engagement levels, helping your nonprofit plan!
Sample prompt: Using this anonymized survey data from the single parents who we work with, pull out the three most common themes in the answers to the question "What childcare services would be most helpful for us to provide to you in the upcoming year?”
Event Management
AI can streamline event planning by generating ideas, outlining steps for successful execution, and helping to create comprehensive plans. For instance, if you're organizing a fundraising event, AI can suggest themes, activities, and logistical considerations to maximize engagement and donations.
Sample prompt: We are hosting a fundraising event with 200 attendees. The event will last for two hours. Please propose a sample agenda that includes: an icebreaker activity (for a group that does not know each other), two keynote speeches from our founders located in San Francisco (which you can find here: link to or paste speeches), a self-service dinner, and networking time. For dinner, include a sample menu with vegetarian options. Also, propose venues in San Francisco that can accommodate the event.
Volunteer Management
Volunteers and nonprofits go hand in hand. They’re like peanut butter and chocolate. But it’s not always easy to match volunteers with the right opportunities. Matching volunteers with suitable roles enhances their experience and your organization's effectiveness. AI can assist in aligning volunteers' skills and interests with your needs, tracking their contributions, and maintaining engagement through personalized communication (Think: thank-you notes and updates on volunteer impact.)
Sample prompt: Please draft sample questions for a survey for potential volunteers. The point of the survey is to understand their skills and interests in order to help us understand what position they will be the best fit for. Our positions include data entry, website design, and program management. Make sure your questions highlight the key differences between these positions.
Conclusion
AI holds immense potential to amplify the impact of nonprofits. By asking the right questions and leveraging all available resources, your nonprofit can harness AI to do more of your mission-critical work.
The journey may seem daunting, but remember — you’re not alone. The tech and AI-powered nonprofit community is rich with shared knowledge, resources, and support.
So, are you ready to embrace AI?