The AI-Powered Nonprofits Coding a Greener Future

Originally published in Stanford Social Innovation Review by Kevin Barenblat and Naomi Morenzoni.

In 1970, Songs of the Humpback Whale was released. The album was one of the first sound recordings of the humpback whales’ song and became an anthem for the environmental movement. It was credited with producing the Marine Mammal Protection Act of 1972 that helped some whale species escape extinction.

At the time, scientists didn’t know what the melodic sounds of whales meant or what whales were saying to each other. But today, a half-century later, that could all change thanks to AI-driven breakthroughs. In May of 2024, scientists published a discovery that just might lead to one of the first opportunities for interspecies communication. With the help of AI-powered nonprofits, like Earth Species Project and Project CETI, we now know that sperm whales have a more sophisticated, expressive, and complicated communication than previously thought.

If the mere sounds of whales led to better environmental protections decades ago, what might be possible if humans could understand the meaning of their communication? Nonprofits decoding non-human communication with AI are just one example in a renaissance of AI-powered nonprofits (APNs) using AI to protect biodiversity and fight climate change.

With the rise of generative AI, debates over its risks and downsides have often dominated the conversation. While companies need to take these risks seriously and prioritize sustainable and responsible AI development, we can't overlook the benefits that AI can bring to the fight against climate change. A 2023 BCG report found that AI has the potential to unlock insights that could help mitigate 5 percent to 10 percent of global greenhouse emissions by 2030.

Fast Forward’s research of how APNs are using AI to fight climate change found a vast range of use cases, including decarbonizing supply chains, tracking pollution, predicting disasters, optimizing sustainable farming practices, protecting biodiversity, and equipping policy makers with better data.

Despite this range of use cases, the number of nonprofits using AI today is limited. According to Salesforce data, 65 percent of nonprofits are open to AI and need to learn more, but only 12 percent of nonprofits are currently using AI in their organizations.

In this article, we’ll explore themes from the research and highlight three case studies of climate-focused nonprofits who are using AI to monitor data, to serve as a coach channeling expertise, and to provide research assistance through better organization. This piece provides practical guidance for both climate-focused nonprofit leaders looking to understand how to harness AI as well as climate-focused funders seeking to learn how to support their AI-powered—and would-be AI-powered—grantees.

Our goal is to showcase the APNs leading the way and empower other organizations to accelerate AI solutions to address climate change.

Major Themes: How Nonprofits Are Leveraging AI for Climate

Fast Forward’s landscape analysis and Salesforce’s observations of the ecosystem through their philanthropic initiatives identified five major themes for how APNs are leveraging AI.

1. Unlocking big data for climate: Much of the data needed to fight climate change exists, but it needs to be made usable by policymakers and decision-makers. Fast Forward’s research revealed examples of APNs like WattTime, Digital Green, and Climate Policy Radar who are making complex data and vast troves of information accessible, searchable, translatable, and usable to a non-data-science audience.
2. Starting small with specific use cases: It might be tempting for a nonprofit to consider all the opportunities to apply AI in its work. But Fast Forward’s research showed the most effective climate-focused APNs started by deploying AI in specific, strategic areas, and then expanded to others. We find it’s best to start small and build data maturity alongside AI capability.
3. Creating tools for collaboration and coalition-building: One of the biggest challenges facing the climate movement is aligning initiatives across sectors and making the most out of our collective resources. We noticed a theme of APNs using AI to create tools that aggregate disparate efforts, build coalitions across languages and countries, and establish a shared source of truth that is open-source and accessible.
4. Leveraging robust existing data sets: We observed the most successful applications of AI come from climate APNs that already have robust data sets that need to be interpreted or analyzed. AI and machine learning (ML) tools are only as good as the underlying data, so funders and nonprofits should be thoughtful about the sources and hygiene of their data. If there are gaps or biases in the data, or the models aren’t trained correctly, AI might not deliver the kinds of insights and impact they're seeking. Great tech tools can serve as the backbone of APNs as they scale their impact and apply AI across their organization. Check out A Better Deal for Data’s learning resources to get started with a data strategy.
5. Using smaller models to drive efficiency and impact: Predictive AI can be just as powerful as generative AI. Many advancements in climate action will come from predictive AI solutions. This is great news given predictive AI typically relies on smaller models, which have a smaller carbon footprint and are more cost-effective than large language models, according to Salesforce research. In today’s AI hype, it’s imperative to remember that advancements in impact will come at the intersection of a variety of innovative technologies.

Three Case Studies of Climate APNs

WattTime: Reducing Global CO₂

WattTime is on the frontier of AI for impact. The nonprofit runs a suite of products that help people, companies, and policy makers make smarter decisions about how to slash emissions and optimize energy sources. WattTime was a participant in Fast Forward’s accelerator and is currently working with the Salesforce Accelerator - AI for Impact, which focuses on AI solutions for climate action and education.

WattTime co-founded Climate TRACE, a coalition of leaders in climate, which gathers and visualizes the most comprehensive and up-to-date inventory of global greenhouse gas emissions data in the world. Climate TRACE uses AI algorithms and a global network of satellites to analyze unbiased, highly granular emissions data from over 352 million emitting facilities. It can pinpoint where emissions are coming from, identify the highest polluting facilities, and determine which facilities are low-carbon emitters with excess production capacity.

Visualization via climatetrace.org.

Leveraging AI, Climate TRACE is unlocking previously unknown climate hacks. For example, many suppliers have excess low-emitting capacity that isn’t any more expensive than other sources. Consider a cargo ship: there is a 50 percent emissions difference per ton of cargo shipping on the cleanest versus the dirtiest vessels. If everyone shifted to the cleaner excess capacity that already exists, WattTime’s data indicate that it could reduce over 1 billion tons of CO₂ emissions per year.

In Fast Forward’s landscape analysis of APNs, Climate TRACE is a clear example of a common AI use-case around monitoring: using AI to continuously collect and analyze data, providing real-time insights and alerts.

Digital Green: Empowering Smallholder Farmers

Small-scale farmers are particularly vulnerable to climate change. Reliant on unpredictable rain and living in lands prone to weather extremes, many farmers are already adversely impacted by climate change. Digital Green aims to solve this by helping small-scale farmers around the world improve their productivity, sustainability practices, and incomes with an AI-powered assistant called Farmer.Chat.

For small-scale farmers in emerging markets like India, Kenya, and Nigeria, agricultural extension agents are crucial lifelines. Similar to a social worker but for farmers, they teach farmers best practices on growing crops. They provide Climate-Smart Agriculture practices, deliver market and pricing information for farmers to maximize income, and help farmers connect with local suppliers. The problem is that many extension agents are spread thin across hundreds of farms and millions of hectares. India, for example, has over 400,000 agents, but the ratio of agent-to-farmer is still 1:650.

The user of Farmer.Chat is the agricultural extension workers, not the farmer. The AI assistant improves the efficiency and effectiveness of the extension worker, expanding their reach and reducing the cost to governments of administering agricultural extension programs. Through this chatbot, extension workers can provide tailored advice directly to farmers based on local weather, soil, crops, and prices. Farmer.Chat is multilingual and multimodal, meaning extension workers can prompt the chatbot with voice, text, and images.

The responsible development of AI has become an imperative for all use-cases. But responsible and ethical development of AI is particularly necessary for products like Farmer.Chat, as it generates answers that directly influence the livelihoods of small-scale farmers. Digital Green has gone to extensive lengths to create an accurate and powerful tool that isn’t trained on all of the internet, but instead on a curated and vetted set of content relevant for small-scale farmers.

Photo via Digital Green.

Built on OpenAI’s GPT-4, Farmer.Chat uses a technique called Retrieval Augmented Generation (RAG), an AI technique that allows companies to automatically embed their most current and relevant proprietary data directly into their LLM prompt. With RAG, Digital Green integrates multiple sources, including its library of over 8,000 farmer-to-farmer training videos (in 50 languages), third-party data sets like reports from local government partners, annotated call center logs, and crop research fact sheets.

Through this technology, Digital Green has provided assistance to 6.3 million farmers, over half of whom are women. Before introducing Farmer.Chat, Digital Green had reduced the cost per farmer of agricultural extension programs by 10x, increased farmer incomes by 24 percent, and increased crop yields by 12 percent. With Farmer.Chat, they anticipate reducing costs of a traditional extension program by 100x. They’re even beginning to see organic usage by farmers themselves, and are developing a go-to-market strategy to serve farmers directly.

In Fast Forward’s landscape analysis of APNs, it’s a prime example of AI as a coach: dynamic AI that emulates human interaction and provides domain expertise in specific fields.

Climate Policy Radar: Equipping Policy Makers

Climate Policy Radar (CPR) builds open databases and research tools so that people can discover, understand, and generate insights on public climate laws and policies. CPR turns thousands of laws and policy documents into accessible and searchable information. Its built-in-house AI models compile and organize thousands of climate change policy and law documents from every country around the world, demystifying the black box of climate laws, policies, and case law into an evidence-based tool for decision-makers.

A powerful, multilingual AI research assistant was desperately needed in the climate space because data on existing climate policies and laws are currently hard to find and analyze. Further, documents and data are spread across diffuse resources, different languages, and inconsistent formats. Learning from mistakes and tapping into best practices is like finding a needle in a haystack, but with Climate Policy Radar’s AI tools, all of this information is available and usable.

Image via Climate Policy Radar.

Climate Policy Radar’s data processing pipelines extract text from thousands of law and policy documents. This makes it quicker and easier to search for climate laws and policies without having to type out precise keywords. CPR’s AI models also organize and associate useful concepts across the texts of thousands of documents. Policy makers compare and contrast how different governments are acting to reduce emissions, responding and adapting to extreme weather events, or regulating various environmental technologies.

With over 6,000 documents that are fully searchable and auto-translated into English and 350,000 users from over 100 countries, Climate Policy Radar is building the world’s largest open knowledge base for climate law and policy.

In Fast Forward’s landscape analysis of APNs, it’s an example of AI to organize: AI to align large corpora of data, ensuring that relevant information is easily accessible, and as a research assistant: AI to sort and analyze vast amounts of data to increase the capacity and speed of research.

Moving From AI Hype to Climate Action

The hype, speed, and risks associated with AI are saturating today’s tech zeitgeist. But amidst all the conversations and speculation about how AI could disrupt everything, there are APNs harnessing it to advance meaningful progress in the fight against climate change. No matter if you’re a small-scale farmer in India or a chatty whale migrating with your pod in the Pacific, climate change is here for us all. With new AI tools, we can accelerate climate action and ensure a healthier planet for future generations, for all species.

Note: WattTime is a partner and grantee of Salesforce.