Ask the Experts: AI that honors your donors and your purpose

Ann Fellman: Okay, doors are open — I’m going to share my screen to let everybody know you are in the right place. This is our wrap-up, our expert panel roundtable conversation on AI: Ask the Experts. Rachel Muir is here with us today — you may have seen her in the chat this week, but now she’s here, and she’s going to facilitate our panel. Rachel, thanks for joining us.

This week has been about AI, and all that it honors in terms of your donors and your organization’s purpose — going through the process of adopting and using AI technologies with intentionality. There have been a lot of great conversations this week, and tons of great questions.

So, a couple of housekeeping items first. I’ll reintroduce myself if we haven’t met yet — hi, my name is Ann Fellman, I’m the Chief Marketing Officer here at Bloomerang. Yes, we’re recording. There’s really no slides today, since it’s a conversation, but you’ll get the recording. At the end, I’m going to push out a poll — if you’d like to raise your hand to engage further with the experts afterward, fill that out. And of course, we’re doing a raffle for a two-hundred-and-fifty-dollar Uber Eats gift card today, so make sure you stay on — we’re a full hour.

If you’re not familiar with Bloomerang, we’re all about helping nonprofits thrive — raising more through fundraising and donor management, and we’re excited about what’s coming this summer with new AI capabilities.

One more foundational thing before we get started: Bloomerang has published its AI principles — you can find them as a blog post and a page on our website. We don’t do AI for the sake of AI because it’s cool; it’s about the mission, and helping you further it. If it doesn’t do that, we’re not going to do it. Fundraising is a relationship business, and it’s important that the relationship stays human — that you’re using your judgment, making the decisions, and staying in control of when and where it makes sense to bring AI into your team’s work. And your data is yours — that came up a lot this week. Your data belongs to you, and we don’t sell or share it. That’s important to understand, and a great thing to ask about with any AI vendor you’re considering.

So, Rachel, the floor is yours.

Rachel Muir: Yes — and thank you! Thank you. I’m Rachel Muir, and this is my dog, Max — he doesn’t bark, so he’s invited to all the webinars, he just sleeps on my lap. I put a link to my website in the chat. I’m so thrilled to be with you all today, and so thrilled to have these amazing speakers who did such a top-notch job all week. This is such an important topic, you all had so many great questions, and you’re welcome to ask more today.

I’m going to kick things off with Loree and Sara, who started us off on Monday. Loree is the founder of Thread Strategies — she was a Case Scholar at Duke University, and we’ve got a lot of North Carolinians in the chat, I love North Carolina. And Sara LaCava Lieberman was with the Innocence Project for almost eight years before joining Thread. I had the gift of getting to know these two wonderful women at GiveCon. They both work at Thread, which Loree started, helping small nonprofits thrive — they did a fantastic workshop, our first time ever having two people tag-team a thirty-minute session together, on AI With Intention.

Loree, you were asked: should nonprofits disclose if something is AI-generated content, for transparency — especially given data showing that it may decrease trust in the nonprofit sector?

Loree Lipstein: Thanks, Rachel, and thanks to all of you for being here. That’s such a great question, and I want to flip it around a little. First, I’d say we don’t recommend just using AI to generate content you don’t have a hand in, and pushing it out as-is — if you asked AI to do something for you, did nothing more with it, and posted it, then yes, you should disclose that it was AI-generated. But I don’t recommend using AI that way.

Instead, think of AI as a partner in your work, the way you’d think of Grammarly helping with grammar or spell-check — a tool that assists you while you, the human, stay in control and are still the one producing the work. That gets you a better product than just dumping a prompt into AI and letting it generate something to post. When you’re using AI that way, as a collaborator, you don’t need a disclaimer saying it was AI-generated, because the bottom line is it wasn’t — it was AI-assisted, but you’re still the one generating it as the expert in your role.

Where you do want your community to know you use AI-supported products is in your data policy and privacy policy — publicly available for anyone to read anytime. That’s how you’re transparent about using AI tools, while still being clear that you’re using them to assist you, not to do the work and get pasted in wholesale. It’s not so much that what you’re putting out is AI-generated — it’s collaborated.

Rachel Muir: I love that — you answered another question in that answer too, thank you so much, that was fantastic. Sara, you’ve got a question: what strategies can persuade skeptical teams that AI is helpful rather than harmful?

Sara LaCava Lieberman: Yeah, thank you. I actually consider myself something of an AI skeptic — a curious skeptic. So I think the first thing is to recognize that people’s concerns are valid; we can’t just say “don’t worry about it, this is great.” There are real concerns around bias and environmental impact, and we’re better off acknowledging those with people rather than trying to argue them out of it.

At the same time, as mission-based organizations, grounding our use of AI in support of our mission — and what it means we’ll be able to do — is what helps bring people along. A lot of what Loree and I talked about on Monday wasn’t just using AI to use AI, but using it to solve a real problem, free up time, and gain insights that have real impact for your mission.

Rachel Muir: I love that, that’s fantastic. Tim kicked things off on Tuesday with rethinking AI — precision over volume. He’s the CEO of Dataro, and an Aussie — what an amazing background, Tim: you were a clinical research nurse, a writer for Concrete Playground, and you have a PhD in cognitive neuroscience.

You got a lot of questions, and here’s one: I know Thread Strategies works with a lot of small nonprofits, and I’ve got a big heart for them too — how can small nonprofits with limited datasets use AI to grow their donor base?

Tim Paris: Thanks, Rachel. This is a question we get a lot, and the answer has changed a lot in recent years. In the past, if you wanted to do something like predictive AI — a really good use case is improving retention, telling you which donors are likely to churn so you can reach out and prevent that, and those donors feel more supported and stay. We’ve done that many times; it’s a great use case. But five or ten years ago, you’d have needed a ton of data and a big data science team to build those models yourself.

These days, there are off-the-shelf tools that do this. I talked about how Dataro does it in my session earlier this week — it benefits from the fact that we’ve turned what those data science teams used to do into software, so it’s plug-and-play. It leverages learnings from other organizations’ donor behavior, so you don’t need all the data yourself.

The best way to think about it: we’ve seen examples of donor churn across broad datasets out there, and we can look for the lookalikes — does this person look like someone who might churn soon? Even with limited data of your own, we can spot the pattern — say, forty-year-old men in this part of the country with this giving history are likely to churn — and give you an accurate recommendation. That’s what’s great about predictive AI in particular: it couldn’t have been done long ago, and now it’s a reliable, off-the-shelf tool.

Rachel Muir: Awesome. I’m going to introduce Meena Das, the CEO of Namaste Data and founder of the AI Equity Project, who lives in beautiful Vancouver, BC — I’ve always wanted to go, never have. She’s a life coach and a first-generation immigrant, very passionate about advancing human-centric data and responsible AI, and she gave a fantastic workshop yesterday introducing the AI equity framework, walking us through all the layers. Here’s one of the questions you got yesterday: how do you get leadership to see all of those layers and the importance of equity?

Meena Das: Amazing, thanks Rachel. Two more things on my bio: she’s also feeling a bit under the weather today, and she also invites everybody to Vancouver — if you’re ever here, I’m taking you out for pastries, I’m addicted to pastries and strawberries. She also has bad internet, so if you can’t hear me, let me know and I’ll type my responses in the chat.

To the question: I want to ask it back to you the way Loree did — how do we know which locks to put on our homes when we’re trying to secure them? The simple answer is we get to know our windows, we get to see where the doors are, what kind of harms exist — we understand our homes in order to know what security system to put in place. It’s the same with ensuring equitable, responsible, meaningful use of AI in our organizations, and making sure leadership is aware of that: we have to start with our own awareness of the inequalities and threats we’re dealing with, and then make it collective.

We have to talk about it in our organization, talk with our staff members about it, and make it continuous — just because leadership couldn’t slate a budget line item for training this year doesn’t mean we shouldn’t next year. Make it collective, make it continuous, and constantly stay aware of the harms possible to your home, your mission — that’s how we ensure leadership stays committed to the layers of inequity we have to be aware of.

Rachel Muir: Awesome, I love that. We’ve got a couple of questions on this next one, plus a live one. Sara — should an AI policy be organization-specific, or is a general nonprofit policy sufficient? Anyone’s welcome to chime in, since there are a lot of questions in the Q&A about building an AI policy — how do I do this, what do you recommend, how do we enforce it, what works, what doesn’t. Sara, go ahead, and anyone else can jump in.

Sara LaCava Lieberman: It’s important to have an AI policy — again, to ensure equity, as Meena was saying, and accuracy, so you really understand how AI is being used at your organization. If you don’t have a policy, AI is probably still being used, you just don’t have the guardrails around it.

You don’t need to reinvent the wheel — like many policies, an AI policy is also a values-based document, and there are great templates already out there. NTEN has a great one, and templates like that are a great starting place; you don’t need to build this from scratch. But you do want to think about your organization’s particular concerns — really understanding your own house, as Meena put it — and how that plays into your policy. An environmental organization might have different considerations than an abortion fund, which might differ from an arts organization.

So we’d really recommend starting from a template, because this can be very overwhelming — “overwhelm” is one of the words we hear most when we ask people how they feel about AI — but definitely look through it with your organization’s specific mission and values in mind.

Rachel Muir: For anyone not watching the chat, Tim just dropped two great resource links in there, exactly along the lines of what Sara was saying — there’s so much great content out there, and it isn’t hard to find; Google it, or ask your LLM to point you to it. Thanks, Tim, for sharing those.

I love that — perfect. Here’s one for you, Loree, that I love — Ben asked: rather than just creating content and writing appeals, what are some examples of innovative AI usage that nonprofit employees tend to skip? Anyone can jump in on this one.

Loree Lipstein: Gosh, that’s such a big question, because coming back to what Sara and I talked about Monday, the best use of AI is so personal to your position, your function, your organization — so skipping something might actually make sense for you. But at a high level, we see a lot of comfort with generative AI tools at this point — your ChatGPTs, your Claudes, your Geminis — and less familiarity with predictive tools, like what Tim talked about, and what Bloomerang is starting to incorporate with Dataro.

I think predictive tools, as a bucket, just feel newer to nonprofits, so they get skipped, or people don’t even realize they exist. Those tools do something very different from a generative tool. So if you’re already comfortable with generative AI for writing, brainstorming, and summarizing, which is great, maybe start understanding what’s available on the predictive side. It does take having the right data — you can’t make predictions off data that’s messy or doesn’t exist, so not every nonprofit is ready for those tools yet. But my overarching answer is to think about the type of tool, rather than one specific function, that’s getting skipped or missed.

Tim Paris: That’s a great call-out, Loree, if I can jump in. You’re right that predictive feels newer, which is funny to me, since — as Rachel mentioned — I got my start in a PhD using that exact predictive technology fifteen years ago. It’s actually been around a long time; it just took a while to get traction, and things like ChatGPT were shiny and new and really started the broader conversation. But predictive AI — specifically the technology called machine learning — has been around much longer, and the upside is that when people discover it, they also discover all the case studies and evidence that it works, from very small organizations to very large ones. It’s one that’s worth checking out. And obviously, if you’re on Bloomerang, that’s something now available to you through the Dataro partnership.

The other call-out on having the right data: if all your donor data is sitting in a spreadsheet somewhere, that’s not the right foundation yet for predictive AI. Once you’re in a CRM — Bloomerang or any other — your data is in a structure, and that structure matters, because the AI knows exactly where things are and it’s clean and formatted. That’s the point at which predictive AI becomes really valuable.

Rachel Muir: Awesome, fantastic. Meena, I have a question for you — actually I want to ask you both this one, I’ll start here. What should a nonprofit ask a vendor before adopting their AI tool, and what does a good answer sound like?

Meena Das: Okay, so — I ask this every time with vendors. I use these two favorite questions not just to evaluate vendors working with my nonprofit clients, but when I’m walking around conferences talking to different vendors too. One: what’s the one feature on your product that you’re incredibly proud of, and why? That tells you what’s important to them — what brings them joy, and whether that aligns with yours. My second favorite: can you show me how you practice responsible AI values or principles on your product? Ask them to walk you through some specifics of how they’re actually living their responsible AI policy, not just stating it.

That gives me a sense of how well they understand their own policies, how willing they are to live them, and how transparent they’re willing to be. Those are my two favorite questions.

Honestly, the first one — “what’s your favorite feature” — might be my true favorite of the two. If the answer comes back pointing to something that could cause more inequity, that’s an immediate red flag for me that our values don’t align. We’re not just looking for good products, we’re looking for partners in this work — so it’s important to understand both institutional and personal values in that conversation.

Rachel Muir: A hundred percent. I’ve got a couple of privacy questions for Loree. One is a note that you recommend an AI policy include how and when it can be used in meetings — you’ve seen people share confidential information captured during a meeting, and post it publicly without checking it thoroughly first; humans need to be in the loop. And the question: for responsible use, how do we identify whether an AI tool has strong privacy policies that abide by HIPAA? How do you evaluate privacy protection?

Loree Lipstein: Great questions. Human in the loop — I don’t know who coined it, but you’ll hear it from anyone talking about AI, and it’s the best reminder that this isn’t something you set and forget technologically; you need to stay part of the process.

The privacy and data policies of AI tools should be readily and easily available to you. If they’re not, that alone is a flag about the platform. If you can’t find it, just ask — I tested this earlier: “Hey ChatGPT, point me to the links explaining your data privacy policies and how they differ across subscriber levels,” or “Hey Claude, point me to your data privacy policies” — and the links come right back, pointing to their publicly posted policies.

You should be able to navigate to this from any product’s homepage yourself, but if you’re having trouble finding it, ask Google or ask the AI to point you to it. If those policies don’t exist or aren’t easily findable, I’d question that organization’s privacy practices — most major platforms make this very clear, show you how to adjust your settings, and give you options to turn certain data-learning features on or off. You want to understand what their policies are and what parts you can control. Anyone not making that available — and trying to hide it — that would be a flag to me.

Rachel Muir: Awesome. I’ve got a big data question, open to anyone — Tim or Sara, maybe. A few people have asked things like: “we’ve got fourteen years of data in spreadsheets, full of duplicates — do I need to clean it up before I can use AI?” And: “how do I get our data into an AI tool to use it?” And you got a question yesterday, Tim: is donor data aggregated or owned by Dataro?

Tim Paris: Great questions, we get these all the time. Let me break it down. On data quality: I think everyone in the industry — and probably everyone in the for-profit space too — thinks their data is poor quality. Especially when you interact with it every day, you see little problems everywhere; it’s the nature of having hundreds, thousands, or millions of records, it’s hard to keep perfectly clean all the time. Everyone will have duplicates — there’s no perfect system that cleans them up in real time for you. Some people have better data than others, but we’ve worked with hundreds of organizations worldwide, and about ninety-five percent of them come in nervous about their data and still get amazing results — or we audit their data against other organizations’ and find they’re right in the middle, totally average. It’s just an impression you get from dealing with your own data every day.

And that’s actually okay — it’s fine to have some duplicates, some missing data, incomplete addresses or dates of birth. My rule of thumb is that if everything’s on a spreadsheet, that’s the point where we don’t know if we can trust the data, because the structure is probably too messy. Anything above that — once you’re in a standard CRM — you’re probably fine, even with only a thousand total records.

The second question is about ownership: who owns the data, where does it go, how is it used. It’s important to call out, especially for Bloomerang customers on this call, that the principles Ann mentioned at the start hold no matter which AI tools are being used — whether built by Bloomerang directly or built with Dataro. You still own that data, full stop. A couple of good questions to ask any AI partner: how are they training their models? In Dataro’s case, we only train on anonymized data — no sensitive information goes into the training process, which speaks to how we operate as a company. And: how do you trust the model — is it a black box? We always try to provide citations, so you can click through and verify the evidence yourself. Those are two important principles to look for in any AI partner. But long story short on ownership: you keep the data, and if you say you don’t want predictions anymore, you don’t get them, and we don’t retain it — that’s the important part.

Rachel Muir: I love it. Meena, anything to add there? I also have a question for you about AI comfort levels.

Meena Das: Go for it, Rachel — let’s do the comfort level one.

Rachel Muir: How do you help an organization that’s a little stressed, trying to figure out its comfort level with AI before committing to anything?

Meena Das: I love that question, I get it a lot in my culture work around AI. I always start with an AI assessment — and when I say assessment, I don’t mean a tech assessment, I mean a culture assessment through language. We’re all at different starting points, even four or five years after ChatGPT, thinking AI is very new. All two hundred-plus of us in this room are somewhere worried about displacement, bias, harm, environmental sustainability — the same concerns, just with slightly different words depending on how we’re choosing to respond.

So before talking about what we can do with AI, I start by capturing the language our staff members use about their fears, their hesitations, their willingness — and look deep into that language to map out which departments and people are the champions, and which aren’t. That’s how you engage people meaningfully as you adopt AI sustainably for the long run. It’s also what makes an AI policy livable — not just another document sitting in a Google Drive or an Outlook folder, but something actually used. Understanding your staff’s language is where I’d start.

Rachel Muir: Awesome, I love that. I’ve got one for you, Sara — how do you stand out, and what prompt should you use, when so many organizations are using AI for social media content?

Sara LaCava Lieberman: Yeah, I think of this as two different questions. Prompt engineering is a modern workplace skill Loree and I talk about a lot, and I want to say more about that. But on standing out — I think that actually doesn’t have much to do with AI itself. You can create a lot of volume with AI, but at the end of the day, the quality of your materials is what makes you stand out, and that’s where keeping the human in the loop matters. Your programs and your impact are ultimately what stand out — AI should be a tool to help you promote that more effectively.

On building a great prompt: what you put in really impacts what you get out, and a strong prompt is usually longer and more detailed than you’d think. We’ve been enjoying a framework by Carlos Morales that he’s shared with nonprofits — a five-step process: think about your goal, the type of communication you’re creating, your audience (both the speaker and the receiver), the specific tone and values you want to convey, any specific information (a generative AI tool doesn’t know what you don’t tell it — testimonials, program-specific statistics), and any constraints around length or formatting.

People often put in a one- or two-sentence prompt, get something mediocre back, and conclude it’ll take just as long to edit as to write themselves — so it doesn’t feel useful. But when you take the time to think through all of that, bringing your own expertise — like knowing you want a PS in your appeal letter, or a specific campaign goal — you get much better outputs with fewer prompts.

Loree Lipstein: Rachel, if I can add to what Sara said — something we talk about a lot at Thread Strategies is how important it is to bring the expertise of your craft to the AI. When we see obviously bad AI output, it usually comes back to the person putting it together not really knowing what they were going for, and letting the final product be whatever came out — that’s when you see the same generic thing everywhere.

Rachel Muir: Slop.

Loree Lipstein: Yeah, the AI slop, as the term goes. It’s really important that nonprofits don’t think of AI as replacing professional development or craft expertise. Since it’s Fundraiser Focus Week, let’s use fundraising as the example — you have to know fundraising best practices to write the prompt the right way, know what you’re getting back, and edit it into something good.

Going back to your original question about standing out: what we see not standing out is when, instead of hiring a fundraiser, an organization decides they don’t need one and just puts a prompt in instead — and you see that same generic output across lots of organizations, whether it’s fundraising or comms. When people know what they’re doing, they can leverage AI in a way you wouldn’t even notice, because they’re going back and forth with it, bringing their expertise — it’s not generated and posted, it’s collaborated with. As you’re thinking about bringing AI into your organization, don’t let it replace training, practice, and people learning their craft — that’s a really important part of getting something useful out of an AI tool.

Rachel Muir: I love that. That brings me to — oh, sorry, go ahead, Tim.

Tim Paris: I just wanted to add one thing — everything there was spot on. I’d encourage everyone to think less about “what should my prompt be” and more about “what should the context be.” Something’s shifted over the past few months where models can handle far more context — it’s no longer just “write me a better email,” it’s “how do I give it all the information.” Like Loree said, you’ve got to train these things like a junior employee — they don’t understand fundraising very well on their own.

But now you can upload entire documents — your fundraising history, how your organization thinks about fundraising, your brand — and put all of that into context. The more high-quality material you put in, the better everything downstream gets. We’re not restricted to small prompts anymore; you can upload many documents that get absorbed intelligently, and get much better results.

Rachel Muir: I love that. This next one I was going to ask Meena first, then Tim: as AI becomes more present in donor communications, how do organizations keep the relationship feeling human?

Meena Das: Oh, I love that question. Tim, I think this is one where you and I should go back and forth.

Tim Paris: Let’s do it.

Meena Das: First — and I get this question so much — I don’t think there’s any formula, gimmick, or trick to making something human. The only way to make something human is to be human. And the only way to be human is by owning what we’re learning — if we make mistakes, we own them, and we give each other grace. That’s how we’re human. If Tim’s learning something new, I’ll give him grace to let him teach me. If I make a mistake learning something new, I want to own up to it. It’s the same with AI — I’ve seen people reject an answer as “not human enough” or “too AI,” and I’ve seen people too quick to say “this should be AI” or “this shouldn’t.” I want us to be more human, not just in what we produce with AI, but in how we show up and talk about the outcomes with each other.

Tim Paris: That’s a really good point. It’s funny — last year I used a slide in all my talks about a famous Chinese robotics competition on the state of robotics, where even the best robots in the world struggled to stand up or shake someone’s hand — there are hilarious videos of them falling down stairs, unable to open a door. It makes you think: what’s the actual benefit of being human? Robots are never going to hug someone, greet them, or have a real face-to-face interaction — there are clear limits to how far AI goes, and I don’t think anyone wants those things replaced, even if the technology someday could.

The more interesting question isn’t just where we can use AI, it’s where we see the real benefit right now. My session this week was really “less is more” — most fundraisers are burdened by so many tasks that they don’t have time to actually be human, take a donor to lunch, have that one-on-one interaction, because they’re buried in emails or appeals or board meetings. The more AI can reduce the busywork that isn’t actual donor interaction, and let you prioritize interactions with the right donors, the more time you free up to do that well.

That’s really where I see AI’s best use case: more time for the right donors, because you’ve freed up time elsewhere and automated some of the steps you didn’t want to do — so you can go deeper on personalized outreach, because you’re not mentally buried under a hundred open browser tabs.

Rachel Muir: I love that. Let’s take that extra thirty minutes we got back in our day and just pick up the phone, call a donor, and say “I saw this, I thought of you.” I absolutely love that, there’s so much joy in it too.

I want to bring in Loree and Sara — this connects to what Tim and Meena just said: what’s the most important thing a fundraiser can hold onto as AI keeps evolving?

Loree Lipstein: It’s what Tim just said — it’s the relationships. Fundraising is about relationships, and robots aren’t going to be the relationship-builders; they might support you along the way, but no human wants to build a relationship with a robot, we want relationships with each other. Sara and I talked Monday about the framework of thinking through what’s taking up your time — and not just what’s taking it up, but what you’d do with that time back, and how to focus on the joy of your job, the uniquely human parts. So the thing I’d recommend holding onto is the people part. That’s why I’m a fundraiser — I love being with people, talking to people, connecting around shared values, helping them connect to something they care about. I wouldn’t give that away, and I imagine most fundraisers feel the same. I really do see AI as freeing up time for more of that relationship-building. Sara, anything to add?

Sara LaCava Lieberman: It’s funny, though not surprising, hearing you say people are what you love, knowing you, Loree — I’m actually a rare breed, an introverted fundraiser, so what I hold onto is the mission. Not that the mission doesn’t motivate Loree too, but that’s what I’d keep at the center — the impact you can make for the people you serve, and the more just world you’re working toward.

The other thing I’d add is: trust yourself, trust your gut as a fundraiser and as an expert. If you read a draft email and something in your gut says “that’s not the message I want to send right now,” listen to that. It can feel like AI has all the answers, but you have the answers — AI is there to free up your time so you can use your own expertise.

Rachel Muir: I love that, a hundred percent. Okay, this one’s a friendly challenge between the two of you — Meena, your framework asks organizations to be thoughtful and deliberate, working through policy and the layers of AI and ethics. Tim, you help teams work faster and smarter. Do you feel like these ideas reinforce each other?

Meena Das: I’ll go first, Tim, because I don’t think there’s anything to argue about here — you’re such a friend. Rachel, the thoughtful, deliberate part of my AI equity framework is the process. Tim’s “faster and smarter” is the outcome of that process — so we’re actually aligned. Neither of us is saying “don’t be intentional” — we have to be absolutely intentional, and I love the idea of slow fundraising. But the outcome of being intentional is that we become smarter, and the outcome of that process is that we become faster. Tim, am I getting that right?

Tim Paris: That’s a great way to put it — there’s a larger goal in both approaches, reached in slightly different ways. Slow fundraising is one way to think about it. AI is good at certain things — let’s not replace what’s really personal and important, and where we can go fast, let’s do it safely and ethically, which frees up all this extra capacity to do the other things much better. That’s where the opportunity is.

Look at any fundraising role — there are hundreds of tasks, and everyone’s spread thin, because most nonprofits are kind of designed that way. There are database analysts who never set out to be database analysts, but had to pick up that role and now the database is a mess — what do you do? The more we can support people in that position, so they can focus on what they actually wanted to do more of — extroverted or introverted fundraising, and I appreciate that call-out — the more room there is to go faster on the places technology can help, and do more of the work that’s critical to what we do.

Rachel Muir: A hundred percent. Here’s a question you got, Tim: if a nonprofit uses a paid AI tool and selects the option that prevents model training on their data, is that data safe to use with AI models while still meeting ethical standards for donor privacy, like AFP’s?

Tim Paris: That’s a question that’ll always come with an asterisk. Taking a step back — if you’re using a foundation model like ChatGPT, Gemini, or Claude, the question is how safely can you use it with your organization’s information. This doesn’t solve every problem, but my quick advice: those platforms have settings that let you turn off using your data for model training. Some do this by default on paid tiers; on free tiers, it’s less likely. If you’re on a paid tier, turn that off — that’s something I’d encourage everyone to do, if your goal is making sure this is safe.

Beyond that bar, there are still more questions — I’d point you back to the earlier answers on building an AI policy and governance, and I dropped a blog post in the chat earlier too, because it’s a bigger, more organization-specific question involving different types of AI and data. That’s a bit of a journey. So it’s not a complete “yes,” but turning off model training on paid tiers is one thing everyone should do by default.

Rachel Muir: I love that — and yes, the recording will be shared. Meena, here’s a question from Michelle: how does an admin team know if their data is clean enough to be beneficial for AI? What do you recommend before taking the step to do an AI assessment?

Meena Das: Yeah — I’ll go ahead and take that, Tim; I’ll say the main thing, and you can add anything you want. Rachel, I’m typing this out — it’s going to be a longer response, so I don’t want to lose it if my internet cuts out.

Loree Lipstein: I don’t have anything to add on the AI assessment piece, since that’s Meena’s area of expertise, but Thread Strategies does a lot of data cleaning with our nonprofit partners, so I can speak to that. There’s more to data cleaning than capitalized names and current addresses — that stuff matters, but really, cleaning your data means thinking about structure: a data architecture, a naming convention. If one year your appeal was called “End of Year Campaign,” the next year it became “Holiday Heroes,” and the year after that something else entirely, it’s very hard to compare those as the same strategy run three years in a row, especially for someone new to the team who doesn’t know those three names refer to the same thing.

We see a lot of data issues that come down to a lack of intentionality in how fundraising strategies get labeled in the data — and that’s not something AI can fix for you on its own. You need to sit down as a team and talk through what your fundraising strategies actually are, what revenue categories you’re tracking, how you want to compare individual versus corporate donors, event revenue versus appeal revenue, what’s annual versus special. AI doesn’t know that — you know it, as the organization, through your development plan.

Once you’ve mapped that out, you can use an AI tool to help finalize the structure or write the policy, but that takes real human understanding of your development strategy first, and some thoughtfulness ahead of the data cleaning itself, so you can compare apples to apples. That level of organizing is often missed when people think about “data cleaning” as just capitalization and missing fields — there’s a whole other layer to it.

Rachel Muir: That’s awesome, you all have shared so much great stuff in the chat. I want to kick it back to Ann, since she may want to run the poll for the two-hundred-and-fifty-dollar Uber Eats gift card.

Ann Fellman: That’s such a party — that’s like a happy hour! So many great questions, and even suggestions about how we might surface or save the answers to these for later reference — I’ll talk to the team, since there are lots of ways we could mine the transcript and publish things from it. Great suggestion.

I’m launching the poll right now for everybody. What I love about these sessions — yes, we’re Bloomerang, but part of our belief statement is helping all nonprofits thrive, whether you’re a customer or not. Everybody’s welcome here to learn from these experts and from each other, and build a sense of community — a lot of your peers in the chat are going through similar things, and it’s just a great community.

So here’s the poll — and the name I drew for the raffle is Joshua Cantu. Joshua, you win the two-fifty Uber Eats gift card! We’ll have Danny Clough reach out to you — hopefully you’ve got a team, some friends, or someone to celebrate with, or just treat yourself to lunch.

So many great things covered today — any other hot takes or questions? Or anything you all want to close with — advice for people, or something you love doing yourself that you’d never hand over to AI?

Meena Das: Rachel, can I answer the “what would you hold onto” question? I love what Loree and Sara said — I want us to hold onto our why. That person we first started showing up to volunteer for, and then ended up doing this work full-time because of. That family at the food bank, that refugee, that immigrant, that person fighting for social justice — that reason to show up still exists out there. AI will keep evolving, that’s inevitable, technology evolves — but our community stays the same, and our why stays the same. I want us to hold onto why we showed up on day one.

Rachel Muir: I love that.

Tim Paris: The number one thing I tell everyone now, in any sector, is that AI adoption is a marathon, not a sprint — and the worst thing you can do is put your head in the sand and assume it’ll blow over. Everyone here is already on the first step, wanting to learn and get new ideas, and there’s a low barrier to start trying things — usually just logging into a tool like ChatGPT and testing it on a new use case. Keep playing with things and finding what works for your organization, because that’s the best way to learn — there’s a learning curve for individuals, but increasingly the bigger benefit shows up at the organizational level, once everyone’s absorbed these tools and the shared context, and you start getting more of an assistant than a one-off tool. That transformation is happening right now — Dataro as a company has been on that journey, and many organizations are too. It takes time, but keep trying, and get your teams adopting these technologies together, because that’s where you really see the benefit.

Rachel Muir: How about you, Loree — anything to add in closing?

Loree Lipstein: I think most of my top advice has already been said — I’ll just repeat that AI doesn’t shortcut expertise. As you experiment with new technology, keep learning your craft, keep learning from others in the space — the more knowledge you bring to the tools, the more they’ll help you.

Rachel Muir: I love that. I’d just echo what Tim said — find time to play around and see what works for you, and keep doing the things that bring you joy that AI doesn’t touch. I still love making slides myself — PowerPoint is my canvas, and I love doing it. Keep doing the things that bring you joy. Sara, anything to add?

Sara LaCava Lieberman: Agreed — everyone surfaced some really great things today, and I think keeping people, mission, and joy at the center of our work is what will keep us on the right path.

Rachel Muir: Thank you all so much for being here today, and for all these insights — you knocked it out of the park with your presentations all week, so much great content, and today was no different. I know a lot of folks are already saying they want a copy of this chat, so I’m sure Bloomerang will figure out a way to share it with everyone. Thank you, Bloomerang, for putting on these Fundraiser Focus Weeks — they’re my favorite part of every quarter, such a blast. I look forward to the next one, and thank you all for your time and for sharing your expertise and hard-won wisdom with this sector we all get to serve.

Ann Fellman: Alright, well — thanks everybody, have a great day and a great rest of your week.