A quick guide for CSR leaders helping nonprofits embrace responsible innovation.
Used wisely, AI can be a superpower for social good—helping doctors catch disease earlier, getting food to people who need it faster, streamlining disaster response, and letting nonprofits stretch every dollar and hour further. No wonder more and more companies are looking to help nonprofits boost data quality, level up AI literacy, and even co-create their own AI intelligent tools.
But used frivolously AI can suck up enough energy to power a small city—just to create what your dog might look like as a human being (guilty). So for companies that genuinely want to harness AI for social good, the question is: how do we build tools that help people without wrecking the planet in the process?
One way is to upskill your tech volunteers on ethical and sustainable AI practices. They should know how to build smart and clean. Here are some specific strategies for limiting the environmental impact of AI tools that might help a nonprofit use or build:
1. Just Because You Can Use AI Doesn’t Mean You Should
AI is a powerful tool (like a chainsaw—amazing when needed, dangerous when misused, and wildly unnecessary for slicing bread). And sometimes, the best solution is… a spreadsheet. Or a hotline. Or a human being using their humanity and common sense. One of the best things corporate volunteers can do to help a nonprofit is to ask:
- Is this problem best solved with AI? Will AI significantly outperform existing methods?
- Does the nonprofit actually have enough clean, ethical, useful data to make it a worthwhile investment?
- AI is only as good as the data it’s trained on. Help nonprofits assess whether their data is clean, complete, consent-based, and relevant. Otherwise, have them start with building better data practices.
- Could the AI unintentionally create new risks—such as bias, surveillance, or excessive emissions?
2. Keep It Lean and Green
Not all AI models have the same environmental cost. A large language model with billions of parameters can emit as much carbon as five cars over their lifetime just from training. But smaller, more efficient models can be equally effective—and far less damaging. Your volunteers can encourage environmentally mindful approaches including:
- Using pre-trained or open-source models
- Fine-tuning pre-trained models, not reinventing the neural wheel
- Running models locally or on green cloud infrastructure
- Using tools like CodeCarbon to track emissions
3. Design With the People You're Trying to Help
Sustainability isn’t just about energy—it’s also about equity. AI tools should be designed with—and for—the communities they aim to serve. Encourage the nonprofits you work with to co-design. Get feedback. Respect cultural context. And don’t scrape data without consent.
Best practices for ethical design:
- Ensure diverse representation in data and design teams
- Involve impacted communities early and often
- Prioritize transparency and explainability
- Avoid extractive data practices and surveillance
4. Measure Emissions and Set Limits
Encourage the nonprofits you work with to track the energy usage and emissions from any AI projects (tools like CodeCarbon can help) and set internal guardrails for what's acceptable.
This past Spring, Salesforce, in collaboration with Hugging Face, Cohere, and Carnegie Mellon University, launched the AI Energy Score, a tool for benchmarking the energy efficiency of AI models. This score aims to provide a standardized, transparent way to measure and compare the energy consumption of different AI models, enabling developers and organizations to make informed decisions about model selection and deployment.
They use this score to measure the energy efficiency of the projects that come out of there Salesforce Accelerator — Agents for Impact program.
5. Fund “Green” Cloud Providers
Encourage nonprofits to choose cloud services that run on renewable energy and have a strong sustainability commitment (e.g., Google Cloud, Microsoft Azure, AWS with sustainability pledges). Or better yet, provide funding specifically for those services.
6. Know When to Review or Sunset an AI Baby
Not every AI project is meant to live forever. Sometimes the most responsible thing you can do is… let it go. Encourage your nonprofit partners to decommission models that aren’t truly serving their purpose anymore. Or that cost more carbon than they’re worth. Or were, let’s be honest, a little overhyped to begin with. Some companies and nonprofits started with some AI experiments to test things out. Don’t let those applications limp along.
Make sure to:
- Continuously audit environmental and social impact
- Track unintended consequences
- Know when to retire or redesign tools that no longer serve their purpose—or that cost too much environmentally
Balance Is Possible
When designed thoughtfully, AI can do both—powering progress without pollution. As technical experts, make sure your employee volunteers understand the trade-offs so that they can best serve their nonprofit partners.
Cheat sheet for your AI volunteers. If you’re building—or investing in—AI for good, make sure it’s:
✅ Necessary
✅ Efficient
✅ Ethical
✅ Designed with community
✅ Planet-friendly
And if they’re not sure, ask. Organizations like Climate Change AI, AI for Good, and CodeCarbon.
