An introduction to the AI Equity Framework

Ann Fellman: Welcome, everybody. I can see everyone flooding in. You’re in the right place. This is Fundraiser Focus Week with Bloomerang. Tell us where you’re at. What’s your location, and what’s your weather? Today, yesterday, and the day before, I was in Dallas, but today I’m back in Minneapolis, and it’s gorgeous out. It’s seventy-five and sunny, a perfect summer day here in Minneapolis, Minnesota. Meena, where are you today, and what’s your weather?

Meena Das: I’m seeing so many beautiful sunny places. I’m also in beautiful sunny Vancouver, BC, although I’m coming with a lot of cold, so I’m remarkably hydrated, sufficiently congested, and it’s beautiful outside in Vancouver.

Ann Fellman: That’s great. A nice summer day. We’ve got a lot to cover, and this is only thirty minutes, so we want to run as tight on time as we can. Thanks, everybody, for joining us today.

This week is all about AI. We started with some foundations and use cases on Monday. Yesterday we talked about intelligent prospecting and targeting, precision versus volume. And today Meena is here with us to introduce the AI equity framework. But before we jump in, a couple of housekeeping items.

Yes, we are recording, and yes, you can get the slides. Remember to stay on, because I’m going to draw a name for a fifty-dollar Uber Eats gift card. Join us tomorrow, too, because we’re doing a two-hundred-fifty-dollar Uber Eats gift card. It’s lunch for yourself, lunch for your team. Someone on my team saves hers up and buys a really nice meal she wouldn’t normally get herself, and I said, I like how you treat yourself. So you can do all kinds of things with an Uber Eats gift card.

My name is Ann Fellman, and I’m the chief marketing officer at Bloomerang. We’re all about helping you raise more. We love to do sessions and weeks like this because they give you access to different experts who can help you strengthen your fundraising motions and better manage your donors and your supporter ecosystem, because at the end of the day, we want to see you raise more. If you’re not familiar with Bloomerang, go check us out later. Don’t do it right now. We feel privileged to help so many nonprofits raise more.

We’re on the topic of AI, and one thing we’ve been talking about this week is that there are a lot of technologies out there that you can use. Bloomerang has AI capabilities as well. I wanted to quickly share that we’ve published our AI principles and how we think about layering AI into technology. Number one, the mission is the metric. There are lots of interesting ideas, but if an idea doesn’t help you and your organization further the mission, we don’t do it. We’ve shot down a lot of ideas.

Number two, fundraising is a relationship business. We want you to decide what you deploy, so we can prepare and draft and help make the work lighter. But at the end of the day, you need to decide, because the relationship stays human.

Really important, too, is your data and your ownership of it. Your data is yours. We don’t sell your data, and we wouldn’t share it without your permission as we deploy these technologies. I think that’s a really good grounding as you think about technology in your own nonprofit. You can learn more at bloomerang.com/ai/principles, so definitely take a look.

Today we’re not going to cover questions. We’ll cover them all tomorrow, so make sure you log your questions for Meena in the Q&A tab, and we’ll get to them in tomorrow’s roundtable discussion. With that, Meena Das, welcome. She’s the founder of Namaste Data and the AI Equity Project. Meena, take it away, and thank you for joining us today.

Meena Das: Thank you so much, Ann. Let me share my screen, and then you can go off camera so we can all see my screen. Can you see it? Amazing, fantastic.

Can I start with two things first? Ben and Pam, you mentioned in the chat window that, Ben, yes, it’s really surprising. We have sun here right now, but it’s going to rain in the afternoon. It’s also going to be a FIFA match today, and I think Canada is playing, so everybody’s very excited here. And Pam, you mentioned in the chat that you’ve visited Vancouver and loved it. Well, this is your open invitation, and in fact an invitation to everybody, all two hundred and eleven people in the room right now. If you’re in Vancouver next time, please let me know. I would love to take you out for coffee and amazing pastries.

I’m Meena Das, and I have thirty minutes to talk to you about something I’m very passionate about. You can see the passion in me holding a tissue with a really stuffy nose. I want to talk to you about the AI equity framework and how it can be useful. Like Charles, who wrote in all caps that he loves Bloomerang, I’m very grateful and happy to say, me too. Bloomerang has been a great partner with the AI Equity Project. I’ll share more about the project after this call so we don’t have to spend time on it now, but Bloomerang and the project have been in partnership for years, and this framework comes out of that.

Before we get into the important pieces, I want to start with a land acknowledgment. I respectfully acknowledge that I’m fortunate to live, learn, and provide my services on the unceded territories of the Coast Salish peoples, including the territories of the Musqueam, Squamish, and Tsleil-Waututh nations. I’m joining from beautiful Vancouver, BC. As a first-generation immigrant, I’m grateful to be where I am, for the land, the water, the sky, the birds, the trees, everything that allows me to be who I can be. I have a consulting practice called Namaste Data, through which I work and talk about data and AI equity, and I’m grateful for the work and the people I get to do it with.

Today is less about me and more about the framework, so let me set some context. The AI Equity Project collects a lot of data from the sector on where we are when it comes to artificial intelligence. What’s the one story of who we are as a sector? This is year three of the project. Last year, two concerning statistics came out of it. First, seventy-six percent of nonprofits are using AI. That’s a number from September 2025, so it will have gone up by now. Fifteen hundred nonprofits participated, so seventy-six percent in the US and Canada are using AI in some shape or form. Second, only fifteen percent have any form of AI policy. That’s concerning. That’s when I realized we had to take the project a step further and create a framework from it.

The AI equity framework is about how you evaluate your AI use cases in fundraising, and how you evaluate whether they have equity in them or not. So my one goal in the next twenty-one minutes is to give you the six layers of the AI equity framework. If we don’t get to all the layers, or you have questions, not only can you reach out tomorrow, but you’re also welcome to reach out outside of these platforms. I promise I’ll send responses back to Ann and everybody.

Let’s quickly establish what AI equity is and why it matters. AI equity is a paradigm and an understanding for ensuring that the systems we choose to use are redistributing power. It’s a lot of work, so let me take you straight to an example. You’ll be getting the slides, so don’t worry about missing anything. Take what feels right to you.

In the example I’ve highlighted in red, a rural nonprofit serving multilingual communities wants to reach more people with limited communications capacity. A simple AI use would be picking up any GenAI tool to translate a survey into three languages. AI equity is a practice, and we know it’s happening when the translation also includes a human. They check that when the survey is translated into Chinese or Spanish or French, it actually keeps the nuance of the language. Board, community members, and leadership understand that simply clicking tools and checking boxes doesn’t mean the language has been translated in the right way, and that trust can be harmed. Community members look at the survey and tell you whether it keeps the nuance. So you’re keeping a human in the loop, not only as a translator but also as a community member. All of those small, significant, humanistic things happening is AI equity for us.

AI equity is really happening when you include the people who are going to be affected by the use, and when you take accountability, not just hold yourself accountable. We can do that when we involve training and resources for staff, and when our practices with AI and data are grounded in consent and community benefit. You’ll get this slide, but I wanted to show you the six layers and spend my time there.

The framework has six layers, and the font is pretty tiny, so when you get the deck, feel free to copy these slides and use them. I want to make this fun. For every layer, I won’t give you definitions or explain why it exists. I’ll give you three questions, along with what an acceptable green answer is and what a bad red answer is. If your answer to a question in your context, your mission, or your job is a red answer, give yourself one point, and count how many red answers you have for each layer. This is not a judgment. It’s a safe space to find out where we are, with vulnerability and honesty.

The first layer is data equity, the most fundamental, foundational layer at the center. We can’t ensure that our AI systems have equity if we don’t give care to our data practices. Here’s how to apply this: if you’re thinking about bringing a GenAI tool into your fundraising tasks, and you want to evaluate whether the use case has enough equity, use this framework. The same goes if you’re building a new platform, a new tool, or an in-house LLM. Pick any example you think is right for you, and then I’ll ask you a question.

For your use case, do you know where your data comes from? A green answer is, “We collected this with community consent in 2022, asking so-and-so. It underrepresents this and overrepresents that.” A red answer is, “We’ve always had this,” or “It came with the CRM,” or “I’m not sure, I think it’s from a past executive director’s time.” If it’s a red answer, give yourself a point. If it’s a green answer, pat yourself on the back. You can also hug your colleague if you’re sitting next to them.

Question two: who is missing from your data? Notice the pattern in the green answers. They have more specificity and clarity, more of the why, the how, and the when. The red answers are more vague.

Question three: can people in your database see, correct, or delete their own data? The green answer is yes, they can when they need to. The red answer is no, it’s not technically possible. These are the questions you can use to evaluate whether your team is ready for AI, whether your tool is ready, and whether your use case is actually equitable. I’m working with one screen, so I’m trusting you here that you’re giving yourself points for your red answers. Be honest. No cheating.

Layer two is transparency and explainability. We’ve set up some good data practices. Now the question is whether the model or GenAI tool can explain what it does to us, and whether we can explain the output to our community. There are two levels of explainability: something to us, and something through us to the community.

Question one: how does your AI make this decision? This is a good set of questions to ask your AI vendors and your tech partners, Bloomerang included. Ask them how the tool works and how they make their decisions, and have them explain the criteria it uses. A red answer is, “It’s proprietary, I don’t know, we use some machine learning, it’s complicated.”

Question two: do people affected by this AI know how it’s being used? Some of that will be taken care of by your AI policy, and some of it will be on you. A good answer is, “Yes, in these cases we disclose.” A red answer is, “No, we don’t want to worry them, it’s all behind the scenes, we don’t mention it anywhere.” Anything vague, unclear, or non-explainable is a bad answer.

Question three: can you show me what data the AI takes in or sees when it makes a decision? Again, this is a good question to ask your tech partners. If they can answer, “Yes, this is what it takes in and how it makes a decision,” good. If you can’t find a good answer, it’s a red answer. For example, “We only see the results, which pull from multiple databases we can track.” Anything vague or unuseful, we can’t do anything with. I hope you’re giving yourself points for the red answers, because I’m going to ask how many you got. The point of this layer is that you can be explained what you’re working with, and then you can explain to your community what you’re doing with your AI tools.

Layer three is accountability. We’ve set good data practices with the data equity layer, and we’re willing to be transparent and explain things. Now, if things go wrong, which they can because this is technology, are we holding ourselves accountable? Here’s a set of questions.

Question one: if this AI makes a mistake that harms someone, who do you talk to? This is a good question for you, your leadership, your board members, and your tech partners. A good answer is specific, direct, and honest. You have a role and a specific person. A red answer is, “No one, we’re not sure, we’d call our main line, I guess.” Anything vague and unspecific means we’re not holding ourselves to the level of accountability we can offer our community members.

Question two: has anyone ever complained or raised concerns about your AI use? Maybe you haven’t gotten into using AI in a very sophisticated way yet, or you’re just starting, so you haven’t had many concerns. If you’re just starting out, take question two as: do you have a place for someone to reach out if they have thoughts or questions? If you do, that’s a good answer. If you don’t, people wouldn’t know it was AI, or no one knows we use it. Again, we’re not offering enough space for people to know that we’re holding ourselves accountable. My leadership is accountable, I am accountable, and my tech partners are accountable for how we use AI with your data for your use cases.

Question three: can you actually stop using AI if it’s causing harm? Any tool is a good place to ask this of your tech partners. Most tools offer multiple AI-based features. If you see harm happening through an AI tool, or you don’t like a feature, can you toggle it off? Can you say, “We still want to learn about it as staff members before we say yes”? I encourage you to ask your tech partners that. And give yourself a point if you’re closer to the red answer. How are we doing on time? I think we’re okay.

Layer four is resource equity. So far we’ve talked about things you can raise with your tech partners and things you need to raise with your community members: data equity, transparency, and accountability. Now we’re looking inward. Do your staff members have access to these resources?

Question one: are any of your staff members paying personally for these tools, or is it through a budget? For equitable reasons, we don’t want it coming out of personal pockets, which is happening in a lot of cases. A good answer is, “No, we fund the tools. We have a process for experimentation. It’s a line item in our budget.” If you’re not doing that, and people are using free versions without AI policies, we’re risking harm to our community. That’s a red answer.

Question two: if we asked your staff whether they have adequate training and time to use AI responsibly, what would they say? A good answer is that you’ve set up trainings. Maybe being here for Fundraiser Focus Week with Bloomerang is part of your training. Those are good answers. A red answer is, “We’ll look at a video,” or “We haven’t really slotted any trainings yet, we’re still deciding, maybe next year.” We need more commitment to collective learning. If that’s happening, it’s a green answer and you don’t need to give yourself a point. We’re measuring ourselves in an easy, gamified way, so one point for each red answer.

Question three: did you budget for AI governance? The policies, the training, the oversight. This is where the equity of access comes in. Are you actually budgeting for that, or is it becoming extra work on an already busy to-do list? I do a lot of AI culture work, and if we’re not committing this budget, we’re losing out on the possibility of making this technology sustainable and useful over the long term. It can be any amount. It doesn’t matter what the number is. It’s about what we’re committing. A good answer is that you’re budgeting. A red answer is, “We’re only budgeting for subscriptions. We’re getting two logins on [inaudible]. That’s all we’ve budgeted.” That’s a good start, and it doesn’t mean you can’t go from a red answer to a green answer. It just means we need more work there.

Two more layers. We can do this. Layer five is codesign and participation. The first three layers were about your tech partners. Layer four was looking inward at your staff. Now we’re doing more community work, going out to the community to codesign and share authority.

Question one: before adopting this AI, did you ask the people it would affect whether they wanted it? This is where you start thinking about how to include your community in the ways you want to use AI. For example, say you want to use a fantastic fundraising tool that has AI features. You’re testing the tools, preparing your test cases, and figuring out a process and an approach for using that tool. You might find a way to bring your donor community and your volunteer community into the conversation about how the tool can be useful. If you’re interested in how we can make that happen, ask in the Q&A. But did you ask? A good answer is, “We found that out in a survey.” A red answer is, “We didn’t want to slow things down, so we assumed and went ahead. Leadership decided it for us.”

Question two: can the people affected by this AI actually say no? It’s not only you deciding, upon seeing harm, that you don’t want to use a feature of an AI tool. It’s also your community member. Say you sent me, as a donor, something that I felt offended or harmed by. I came to you and said, “Can you not use my data in your AI feature?” Use that example to go back to your tech partner and say, “One of our donors doesn’t want to be part of this particular feature. How do we make that happen? How do we hold consent as being of the utmost importance and care, more than how we use AI?” That’s a green answer. A red answer is, “We’re not sure. We’re not the decision makers. We don’t have control over the tool.”

Question three: who was not in the room when you decided to use AI but should have been? I cannot stress enough how important this question is. A good answer is that you can name the groups: the frontline staff, the people who do the intake forms, the volunteers who work over the weekends collecting what the community needs at the food bank. Anything specific is a good answer. A red answer is one that tries to protect what doesn’t need protecting and should frankly have been changed by now. So if you can name a specific set of people who should have been in the room but weren’t, that’s a good answer.

The last layer is power and self-determination. Same thing again: hopefully you’re giving yourself a point for each red answer.

Question one: who ultimately controls the AI that affects your work? You, a vendor, or a funder? Some of these questions may feel like they need a discussion, because they’re beyond a binary green or red answer, which is very fair. I’d ask you to have a proper discussion with your colleagues, your board, and your leadership. But a good answer is that you know where your data is coming from, what data privacy laws apply on top of it, and what data equity values you place on top of it. A red answer is feeling that you don’t have control over your data, which we don’t at many points. But I want you to start knowing that we always have a choice, and we can take steps accordingly.

Question two: is this AI helping you build community power, or just making your operations more efficient? This question is about how and why you evaluate your AI tools with respect to your mission. Is the tool actually helping you or not? If it is, and you have a clear answer, that’s a green answer. If all you can say is “it’s going to save us time,” then we still have to dig deeper to figure out how it’s human-centric and how saving that time helps us maintain good, solid relationships with our community.

Question three: if this AI is working well for you but harming the people you serve, what would you do? That’s a big if, and obviously we don’t approach it with that big if, but let’s plan for it. If something goes really wrong, are you willing to step back? Are you willing to hold yourselves collectively responsible? How would you fix it? Let’s make a plan for that, and let’s share that plan with the community. Let’s share with the community how you’re choosing to use AI.

Let me open the one picture with all the framework layers. I hope you’ve been giving yourself a point for each red answer. Let me look at the chat window. How many red answers did you get? We have six layers with three questions each, so it can go up to eighteen points. Kevin says fifteen. Caroline, twelve. I so appreciate the honesty. Thank you. Juliana is at seven. Who else? Anyone between fifteen and seven? Thanks, Charles. Heather says ten. Ruth says twelve. I love the honesty, and I love that you’ve all been following all the questions. So far we have two sevens. Thanks, Noah. Anyone else? Twelve. You are all fabulously honest, which is great. I think this room is somewhere between ten and twelve. Fantastic. That is very honest. Ann, are you looking at this?

Ann Fellman: I am. This is great. What you hit on is that you’re going to have red answers or yellow answers, and the point is to be intentional in these different areas and topics and ask, what’s right for our organization? What are we comfortable with? It’s a really good framing for having intentionality behind how, where, and when you use AI and how you train it.

Meena Das: I love that. True, Ann. And Ruth, I want to acknowledge your question. You’re asking what is meant by harm. We’ll take it to the Q&A tab, and if we don’t get to that question tomorrow, I’ll write it back to you, and please feel welcome to share it with Ruth.

Ann Fellman: Great. Log those questions for Meena in the Q&A tab. Great commentary. If you’re interested in learning more and want to connect with Meena, you can raise your hand here. You can also schedule time with Bloomerang if there’s something you want to explore. I saw somebody say they’re new to Bloomerang and there’s always so much to learn, and we’re always shipping new capabilities. So let us know. Raising your hand is a great way for the team to reach out to you. This was fantastic. Meena, thank you so much.

Before we go, Steve Dabrowski, you are the winner of today’s fifty-dollar Uber Eats gift card. Congrats, Steve. Come back tomorrow, everybody. We’ll have Meena, Tim Paris from Dataro, and Loree Lipstein and Sara LaCava Lieberman joining us for a great conversation, all facilitated by Rachel Muir, who is fantastic. It’s a roundtable discussion about all things AI, and about deploying it in the most intelligent ways, no pun intended, in our organizations, being thoughtful and intentional. We’re all busy, so how can we do this in a way that genuinely respects our relationships with our supporter community? There’s also a great opportunity here to level the playing field, because our teams are often stretched thin, and there are a lot of ways new technologies can help us. Tomorrow’s a full hour, longer than today’s thirty minutes, so we’ll see you then. There are plenty of questions already in the question bank. Great discussion. Please join us tomorrow. Meena, thank you so much for joining us today. Have a great day, and see you soon.

Meena Das: Thank you, everyone. Bye.

Ann Fellman: Bye.