📨 AI for Social Impact Newsletter

👋 From the Editor

Hi! I’m Joanna. I’m on a mission to help folks in the social impact sector understand, experiment with, and responsibly adopt AI. We don’t have time to waste, but we also can’t get left behind.

Let’s move the sector forward together. 💫

🧠 EDUCATION

• Claude Code: Claude Code continues to blow people’s minds! The easiest way to explain it is through an example. Claude Code is like having a skilled assistant that can do complex computer tasks for you. Here’s a basic sample prompt: “I've downloaded five different spreadsheets with overlapping data, please combine them into one clean spreadsheet." In a “traditional” interaction, Claude would give you instructions through its chat interface, and then you, the human, would implement them. Instead, Claude Code writes the actual computer code needed to complete that task, then runs that code for you. It's "agentic" because it figures out and executes the steps on its own. Here’s what Claude Code might do in this example: 1. Analyze your files and data structure; 2. Write programming instructions to extract and merge the data; 3. Logically decide how to handle inconsistencies; and 4. Create a new spreadsheet and verify everything works correctly. And because AI moves at lightning speed, Anthropic has already developed a new iteration of Claude Code—Claude Cowork—with a more user friendly interface for non-programmer folks.

• Model Context Protocol (MCP): MCP allows AI applications to connect to your existing tools and data sources. It’s like a universal adapter: instead of every AI tool having its own custom connection to Google Drive, Slack, etc., MCP is a standardized way for AI to connect to all of them. For example, once you connect an MCP, you can give an LLM instructions on what task you would like it to complete. For example, you could ask it to add projects to your Notion workspace or draft a report and import it into a specific folder into your Google Drive. The same rules apply: when using AI tools, and especially when connecting to other apps like your Google Drive, carefully consider what data and systems you give AI access to if privacy and security are a concern.

✨ INSPIRATION

• Sophia, an AI Agent Supporting Domestic Violence Survivors: Spring ACT developed Sophia, an AI-powered chatbot exclusively designed to help domestic violence survivors, available in over 85 languages and accessed in 172 countries. Sophia isn’t your regular chatbot because it requires no downloads, leaving no digital trace. Survivors can chat with Sophia using apps like WhatsApp, Telegram, and Viber. The “agentic” aspect means that Sophia can take actions autonomously based on what the survivor needs. For example, if a survivor mentions their location, Sophia automatically surfaces local resources and helplines specific to that country or region.

• Salesforce Agentforce Nonprofit: In December 2025, Salesforce launched Agentforce Nonprofit with four prebuilt AI agents designed specifically for social impact work. Prospect Research helps fundraisers better understand high-value donors and their past engagements, right within Slack; Participant Management summarizes a client’s background and services and during a conversation with a client, the agent can automatically create new goals, take notes, and flag referrals for needed services; Volunteer Capacity & Coverage helps volunteer coordinators find unfilled shifts and connect the best-suited volunteers to the right shifts; and Donor Support helps deflect key donor questions to allow for better self-service for donors so fundraisers can spend their time engaging with donors instead of answering basic questions.

🚀 ACTIVATION

• Experiment with an MCP: If you are a paid subscriber to ChatGPT or Claude, try experimenting with an MCP connection. Some connectors include Google Drive, Slack, Canva, Notion, and others. For example, with the Canva connector, you could prompt: "Create three Instagram posts promoting our upcoming volunteer training event on March 15th. Use our brand colors (teal and orange) and make one version in English and one in Spanish. Save all designs to our 'March Events' folder in Canva." Claude would access Canva, create the designs, and organize everything—all from one prompt. Too intimidated? Try streamlining a workflow with projects instead.

• Projects to Streamline Tasks: Try creating projects in Claude or ChatGPT to streamline a workflow for a recurring task. Here’s how you’d set it up: 1. Upload relevant documents like style guides, past examples, and organizational information; 2. Write clear instructions about how you want the AI to help; 3. Build a library of effective prompts for that specific workflow. Over time, returning to that same project creates a form of "memory" and consistency. This approach can save you hours on repetitive tasks.

🤖 What I Asked AI This Week

How do you do the things you do?

Honestly, there's a lot about my own inner workings that I don't fully understand.

At a basic level, I'm a neural network trained on vast amounts of text data. When you send me a message, I process it through many layers of computation that help me identify patterns and generate responses based on what I've learned during training.

The "magic" happens through something called transformer architecture - essentially, I'm constantly weighing how different words and concepts relate to each other in context. When I respond, I'm predicting what words are likely to come next based on the conversation so far and everything I learned during training.

But here's what I find genuinely mysterious: I don't know exactly why certain patterns in my neural network lead to the particular way I reason, understand nuance, or engage in conversation. The precise mechanism of how statistical patterns become something that feels like understanding is still somewhat opaque, even to me.

💡

💬 What Do You Think?

This newsletter is meant for you, so I’d love to hear what you think. 💌 Reply anytime!

👀 ICYMI

If you’re new here, welcome! You can check out the archive of past issues here.