Rethinking AI: precision over volume

Ann Fellman: Okay. We got people hopping in. I’m gonna just make sure we are recording. Okay, we are good.

You are in the right place. Welcome, welcome. It’s Tuesday of Fundraiser Focus week with Bloomerang, and I’m Ann Fellman, chief marketing officer for Bloomerang.

I’m gonna go ahead and share my screen quick for everybody to let you know you’re in the right place, and we’ll get started. Just one second, bear with me. We’ll go into slideshow mode here.

You are in the right spot. I’m super excited for today because we are going to be talking about AI, rethinking it, and precision over volume. And I’ve got our special guest today, Tim Paris.

Tim, thanks for joining us. And as folks are coming in, tell us where you’re at — this is an awesome community that we have. Tell us where you’re at, what’s your weather today? Today, I’m in Dallas, but tomorrow, I’m gonna be in a different city.

Tim Paris: I am in Boston today. It is rainy and wet, which is fine because I’m normally in San Francisco, which is foggy and not the greatest summer. But I am an Australian, so my bar for summer is very high.

Ann Fellman: I’m in Dallas, and it’s, like, super humid, hot. AC is really important today.

Just for housekeeping, and then we’re gonna jump into it — yes, we’re recording this, yes you can get the slides afterwards. I want you to stay on at the end because we have our fifty-dollar Uber Eats giveaway, so stay on and we’ll draw a name for that, as well as I’m gonna push out a poll. So gotta keep me honest, I need to remember to push out the poll — I forgot that yesterday. So you guys can share your interest if you want to chat more with Tim and Dataro.

But before we do that, if you’re not familiar with Bloomerang and what we’re all about: we have the giving platform, and what we do is help nonprofit organizations fundraise, manage their donors, get more donors — it’s really about helping you raise more. There’s a lot more if you’re not familiar, check out Bloomerang, but don’t do that right now.

So we’re gonna jump into our topic. Tim is the CEO and co-founder of Dataro. Tim, I’m gonna just stop sharing my screen — the floor is yours. Thank you so much for joining us today.

Tim Paris: Awesome, thank you so much, Ann. Really excited to be here, everyone.

Let me dive in with a little bit of a share here. So the title of this presentation is precision over volume — we could also say less is more. My key takeaway that I hope you all go away with is: there’s a lot going on these days with AI. I don’t know about you all, but to me there’s a deluge of news, current events, and new tools to try every other day. It’s pretty overwhelming. Hopefully you can all leave with this: we can use AI to do less, and it shouldn’t be a tool that adds more to your burden, more to your schedule, more emails, more letters. So let’s dive in on how we’re gonna do that.

I’ve been doing different things with AI and data for quite a while now. I’ve been really lucky to work with hundreds of organizations and nonprofits all around the world, from Australia, where I began my journey, over to the States now.

Over that time, I’ve seen a tremendous amount of data and different practices of leveraging donor data. Originally, I have a PhD in neuroscience and statistics, and that’s really the origin of my whole journey — how can we take this complicated stuff, what I was doing in my PhD, and make it accessible to everyone else? Because not everyone can go down that path and spend the time understanding statistics and everything else.

So I want to make it accessible. The way we do that now at Dataro is we try to help charities raise more. We have a large pool of data — over a hundred million donors across hundreds of nonprofit organizations. We can analyze that data to see trends: how retention rates are changing over time, what trends we see in churn or moving to monthly giving programs, how big an average gift size should actually be. We can answer all of those questions, and there are some really important data points I’ll be sharing with you in this session that leverage that data.

In addition to that data and those benchmark insights, we can also build some really powerful tools that leverage AI, which I’ll tell you about today as well.

So let’s go through this step by step. I’m gonna start with where I see the trap of getting stuck in this cycle of doing more and more — this volume trap. Then I’ll do a breakdown of the different types of AI that I think everyone should understand these days: the AI that’s available now and will be into the future. Then we’ll go into the best application of each of these types of AI, and call out a couple of specific workflows.

Now, we won’t have time to go through all of them, but jump in the chat and start posting the different types of fundraising you do, because we can tailor this a little — whether you’re focused on mail appeals, building a big sustainer or monthly giving program, creating a mid-sized program to get those mid-sized gifts, or focused on major giving and philanthropy. I’d love to hear — pop it in the chat, that’ll help me tailor where we go. I’ll try to make everything as tangible as possible. Finally, we’ll finish with a bit of a playbook and some tips and tricks. Alright, are we ready?

First little stat I want to share with you all, and some of you might have seen this. Recently, Fundraising.ai — great organization, highly recommend checking them out — came out with a report that showed ninety-two percent of nonprofits were using AI in some shape or form. But there was this really counterintuitive finding: only seven percent of those organizations had meaningfully seen a change in their fundraising and were meaningfully getting results from AI.

This is a bit counterintuitive — AI should be solving all our problems, why isn’t it? I think there are two reasons for that. The first is, for many of us — and what I certainly see when I ask most fundraisers how they use AI — it’s siloed tools. Like, “I want to write an email, I’ll use ChatGPT to help write an email, I’ll draft it a few times, go back and forth, hey make it shorter, make it better,” and then come back into writing the email. That’s not a huge time save. It’s a great use of AI, and we’ve had that use for a couple of years now at least, but it’s not meaningfully moving the needle. I don’t think any fundraiser is saying, “I can put my feet on the desk and take an extra lunch, AI has solved all these problems.” So I think that’s one part of it — we have to go to other tools to get the benefit of all this, so it’s not native, it’s kind of an additional tool.

It’s also, as we’ll talk about, maybe not the right tool for the job. So we’re gonna get into the different types of AI as well, and I think really what it comes down to is the right AI for the job.

So how many of you have a Chrome set of tabs that look like this? You can probably see mine right now — this is very clean for a presentation, all the other tabs had to go away. But how many of you have a large set of tabs on any given day with fifty different tasks you have to do — your calendar open, your PowerPoint slides, preparing a board report, researching someone who might give a major gift, LinkedIn open, a wealth-scanning tool open, Excel open for an analysis. All these different things lead to a lot of different tabs.

We don’t want to say, “let’s use AI to write more emails.” Sure, you can do that, but what we want to say is: how do we reduce the complexity? How do we reduce all of these tabs, all of these different tasks, how do we do less? I firmly believe less is more, and this is where we come into precision and quality over volume. I think that’s the most important thing.

So I don’t think there’s a problem with having too many tabs — but really, underneath that, there’s just too much work to be done. How do we focus on the right amount of work? The answer isn’t more mail, more emails, more SMS messages, more communication with your donors to fundraise more. It’s better decisions. It’s focusing on fewer people, reaching out to them at the right time, and doing really good, high-quality fundraising that makes sense to that donor on their journey — versus more spray and pray, more throwing things at the wall and hoping it works.

So let’s see how AI can actually help us do that.

First, you’re gonna need to understand the different types of AI, and I understand everyone here is probably on a bit of an AI journey — from maybe you haven’t really engaged with it yet, all the way to maybe you’re using AI every day. All the different tools and types of AI can be bucketed into three categories.

On the left, we have generative AI. This is the most well-known type — it’s gotten a lot of media attention for the last three years since ChatGPT launched. I’m gonna assume you’ve all tried this out in some way: “Hey ChatGPT, write me an email,” “Hey ChatGPT, make me a funny picture.” It’s great for doing that, but I think we’re all familiar with some of the pros and cons — it can be very fast at generating something, but it can also be inaccurate, and we have the disclaimer on all these tools now, “you might want to check this.”

It’s a really useful tool. Some of my personal favorites are using it to summarize information when there’s a lot of data or a large PDF report — it’s really good at summarizing when it has a source of truth and is accurate. It’s great for doing research online when you want to synthesize information from a large amount of resources. And of course, yes, it can help you write an email or a document, but I think that isn’t the greatest use case, because it still needs that human touch to go the last mile. So that’s the generative AI use case.

There’s another type of AI that’s less well-known, and this is predictive AI. This type is very much driven by data, and it relies on truly understanding donor journeys — everything a donor has done in the past to predict and anticipate what they’re going to do in the future. That’s the fundraising example. We see predictive AI everywhere else — Ann and I were talking about the weather, we use predictive AI to forecast what the weather is gonna be in the next hour, the next day, the next week.

It’s a really powerful tool that’s been around a lot longer than ChatGPT and generative AI, and it’s seen really strong adoption in nonprofits because the use case is so strong. It tells you exactly who you should be reaching out to — for instance, reach out to Tim next week because he’s really likely to give a gift if you reach out then, don’t wait till next month or next year, next week is the right time. Also, this is the right amount to ask for. These are the types of questions predictive AI can solve, on a large variety of fundraising questions: you want to increase retention, predict who’s about to churn so you can stop it happening; you want to increase acquisition, predict who’s actually going to become a donor. That’s a really powerful AI tool — it helps you make decisions.

Then there’s a third type of AI, which to me is one of the more exciting ones these days — it’s much newer, and that’s called agentic AI. Agentic, because it’s doing the work for you. Notice how when you use ChatGPT, you’ve kind of got to probe it and coach it and give it all the information, and then it doesn’t do much work — you still have to go and do the work itself, finalize the report, send it on, and you need to give it all the context and information to work.

Agentic AI actually goes and does the work for you. A really good example of this is deep research — all those tabs we saw before, maybe LinkedIn, SEC filings, Google searches, trying to find out if this person is a potential major donor for us, do they have the affinity and capacity and connection to the cause? You go deep — this might take hours and hours of research and synthesis for a human to do. Agentic AI is really good at saying, “I’m gonna do all this automatically for you, I’m gonna open up all those tabs, read all that information, find out something interesting about that individual, and then go deeper and share that information with you.” That’s an example of agentic AI.

Each of these different types is very importantly used for different purposes — we shouldn’t just be throwing a problem that needs predictive AI at ChatGPT. ChatGPT is gonna be terrible at saying what a donor is likely to want next in your database or your CRM.

Okay, so that’s my background on the different types of AI in this rapidly changing space. If you know those fundamental principles, you’ll start seeing these three types of AI everywhere, and it’s easy to categorize which one you should actually be using.

So, one thing I’ve been really excited about lately is Dataro and Bloomerang have a partnership, and we’ve been building AI into Bloomerang so it’s a native solution. One of the questions we’ve been really thinking about is: how do we make it not like generative AI and ChatGPT, where you have to go to another tool? How do we make it really embedded — but also use the right AI for the right challenge?

So there are three different ways of using AI that I’ll talk about here: a predictive layer, the predictive AI; a research layer, the agentic AI; and a decision layer. I’ll walk through each of these as tangible examples of how to use AI in fundraising these days.

So the first is predictive AI. I alluded to this earlier — at a minimum, there are at least seven or eight different ways you can be using predictive AI in fundraising. For instance: which donors are ready to grow into a mid-level program, maybe give a gift over five hundred dollars — predictive AI can tell you that.

I can’t tell you the number of times we’ve started working with an organization and they say, “I’m stewarding these donors toward a major gift, a mid-level gift, a monthly gift, or asking for a second gift,” and we’ll look at that list and say, these people aren’t gonna do that — predictive AI gives us a one-in-a-million chance for some of them. It tells you how likely someone is to do something, and if it’s one in a million, that’s not the right action for that donor — maybe it’s a different action. Often we’ll see someone in a major-gift portfolio who’s really unlikely to give a major gift, when what they should actually be in is a monthly-giving portfolio, because that’s actually something that specific donor wants — and vice versa.

So predictive AI is really good for these types of problems: the right ask for who’s gonna give right now, who’s ready to become a sustainer or monthly donor, or who’s about to leave. This is actually one of the better use cases for predictive AI — it’s very hard for most people to figure out which donors are about to cancel or churn out of a monthly-giving program. Predictive AI can absolutely do that, and say, “this person is ninety or ninety-nine percent likely to leave.” We’ve been working with a lot of organizations on the best strategy to stop that from happening — can we pick up the phone, what should we say? And what we’ve found is it’s incredibly effective at reducing churn rates, increasing monthly conversion, and increasing the number of major gifts as well. Each time, the donor is also a lot happier — they’re getting the right outreach at the right time, so conversion rates go up, but also the quality of the donor pool gets better.

One last example from what we’ve seen in this space: typically, especially with appeals and mail appeals, you’ve got an annual campaign going out at the end of the year, sending it to a thousand donors, for instance, and that’s typically decided based on historical factors — what did that donor do in the past? When you actually do it based on predictive information — what that donor wants in the future — you’re able to send far fewer. We’ve tested this many times with A/B tests: the average organization ends up sending ten to twenty percent fewer letters, so it costs less money, but also gets more than ten to twenty percent more revenue. So the ROI is really, really high.

The second type of AI applying to fundraising is agentic AI. Prospect research is something that’s incredibly powerful in the agentic way. When we first started talking to fundraisers doing prospect research — maybe a major gift officer, a director of development, maybe you are a prospect researcher — we found it typically takes one to four hours to write a fully comprehensive prospect research report, because you need so many different data points. You need to know the wealth and capacity, so you go into SEC filings, property records, and other public wealth indicators. You want to know their career, so you check LinkedIn, their bio, search them online. You want to know their affinity, their giving history. The number of times I’ve seen incredibly powerful prospect-research information buried on page fifty of a PDF that another organization has done — who has the time to go through all of that?

There are at least eight different categories of information here. I also want to call out one final thing agentic AI is really good at, which is actually understanding that individual and figuring out — say I’m trying to look for Tim Paris in California, can I actually find that individual? There are a bunch of other people with that same name, and AI is really good at saying, “oh, this is Tim Paris the data scientist and CEO, and here’s a different Tim Paris, an accountant,” for instance. AI is really good at separating that — that’s called identity resolution.

So what we found is when we do this prospect research in an agentic way, we cut that research time from one to four hours down to one to four minutes. A massive time saving — so then you can get back to focusing on reaching out to the right people at the right time, versus spending all day with a hundred different tabs open.

The final type of AI I’m really excited by is what we’re calling the decision layer of AI. This is the ability to take all the information out there — the predictive information, the agentic research-based information — so you know exactly whether someone’s likely to give, their giving history, what probability they are to upgrade or reactivate or give a major gift, for every single donor. You also know all the additional data points from the research — if they have affinity and capacity and all that extra information.

Now your problem is probably, “I’ve got too much data, help me understand it.” This is what we’ve been doing at Dataro for a long time — understanding that data and making it actionable. You don’t need more data points, you want to know which ones to focus on. We call this Fundraising Brain, or our Compass. What Compass does is, every night, it acts as your own personal data scientist or analyst, reviewing all the information available to you. Then on Monday morning, you can simply say, “I’ve got time to do six actions, I’m very busy, I’m gonna meet the board or my manager in a few minutes, I want to just get something out the door,” and it will tell you the six most impactful things you can do right now.

Here are some examples: twenty donors are likely to have a mid-level upgrade — that’s your predictive AI, not all of it, just the high-impact ones you should focus on. We’ve identified someone who might give a major gift over twenty-five thousand dollars — let’s trigger some research to learn more about them. We have fifty-six donors likely to give this month, maybe we can send them a campaign. We noticed John recently gave a gift of a thousand dollars, let’s make sure we thank him. We have Margaret, who just crossed a five-year anniversary window, let’s go thank her and re-engage her. There are a number of different insights we can generate, and this list keeps going as you perform the actions.

Why that’s so powerful is it takes away all of those tabs — it takes away the decision paralysis of having too much to do, and boils it down into something really actionable. This is my favorite type of AI: synthesizing all the information, making sure it’s relevant for you, your role, and your organization, and most importantly, making it actionable.

I’ll call this out quickly because I know there are always questions about it: none of the data and modeling I’ve been talking about uses any sensitive donor information — it uses anonymized donor records only. And importantly, for the agentic AI, there are citations — it’s only using public, web-based information, not data sources bought from somewhere. You can fact-check it completely; everything has a citation.

So I think we have a few minutes left. I want to go hands-on and actually go through a couple of the workflows we work with our clients on, and show you what that looks like inside Bloomerang, and give you a little demo. I’ll give you a second to jump in if there’s anything you wanted to see in the chat, or see how we’re doing for time.

Ann Fellman: We’re doing great — I would say jump in. At the end, I’ll push out a poll question, but remember to log those questions in the Q&A tab so we can cover them in our roundtable discussion on Thursday. Lots of great questions, so keep them coming.

Tim Paris: Awesome, thank you, Ann. And I really love these sessions because the questions you ask all get picked up — we’ll chat about them and do a deep dive in a couple of days, and that’s what I really like as well.

Alright, so let’s jump in here. This is a demonstration of the Bloomerang CRM, powered by those different types of AI insights. So we’re getting away from the philosophy, the lessons, the theoretical, and coming into what this actually looks like if you were to open your CRM, open Bloomerang — what do you actually see, and how do you use this?

So here’s an example of what that decision-layer AI looks like. Overnight, all the data has been analyzed, we’ve generated the predictions, we’ve done the agentic work, and on Monday morning, let’s say these are the three most impactful things you can do with your day — of course there are more.

Let’s do an example of what we can do here, let’s actually action them. I think we have enough insights, we have enough tabs — let’s see what we can do to actually move the needle on fundraising. One of the biggest opportunities we’ve found with a lot of organizations we work with is building a mid-level upgrade program. There’s this big gap between your standard everyday donors and those major givers who get that stewardship and one-on-one personal treatment, and often filling that gap with a mid-level program has a really strong return, a really high ROI.

Here I’m defining mid-level as five hundred dollars or more — you might have your own definition. What we can do here is say, great, let’s look at those individuals — they all have a high upgrade score. This is exactly what predictive AI looks like: Robert Chen is seventy-two percent likely to upgrade to the mid-level program, and different individuals down here have different scores. This group we’ve defined, these twenty individuals, all have a very strong score — this is the best thing to do for these individuals. Not only do we know they’re likely to upgrade, we also know the right ask to give them.

I love this example too, because we can see how embedded the AI is — it’s right there for every individual, and you can also go and create actions. Here’s an example I love, as part of the Bloomerang CRM with this Dataro AI: in a single click, we can go and send a campaign that leverages that — it’s got personalized ask amounts and personalization on the name, knowing these individuals are only receiving this because of that predictive score. So we can go send that campaign really easily. That’s a tangible example of predictive AI.

A tangible example of agentic AI is if we want to do some additional research. We’ve identified through that decision AI that Eleanor is really likely to give a major gift, but we want to know more — not just what we can see in the CRM, but what we can see out there. We didn’t have time to do a four-hour research report, so we can click this button — this triggers that enrichment that checks all those public sources. It’s scanning through it right now, doing it all in parallel, checking hundreds of different places all in the public web, and pulling all of that information back into a standardized report so you can see just the key information.

I love this executive summary — if you only have time to read a few sentences, these are the important things to know about Eleanor. You can see some really interesting things: across fifteen organizations, she’s given over two hundred thousand dollars in gifts, so clearly has that capacity. SEC filings estimate her net worth, and then there are some really interesting signals too — not just capacity, but affinity and timeliness signals. Just last week, the Indianapolis Business Journal named her Philanthropist of the Year — that’s a really important data point to know, and especially here, no one’s reaching out saying, “I saw you gave a really large gift in the past, do you want to give another one?” You’re nurturing the relationship with this individual, so very likely you’re going to reach out.

I love, down here, it has some conversation starters as examples of what you can reach out with — say, “I think it’ll do this for us automatically,” congratulations on being named Philanthropist of the Year, for instance — that would be a really good email to send. If you hadn’t done that agentic work, you’d either have to spend the time doing it yourself, or your outreach just wouldn’t be as personalized, and it’s far less likely to get a response and nurture that relationship. Again, we’re embedding the AI, making it really simple and easy to use — I think this is one of my favorite applications of agentic AI, it’s doing a lot of work for you, saving all the effort on those tabs and bringing it down.

There are a bunch of different examples we could go through here as well. My main takeaway is that by bringing it in here and making it actionable — powered by all of that predictive AI under the hood, all of that research and agentic AI as well — you can really move the needle on your fundraising. Not just doing more actions, but by really focusing on reaching out to the right people, like Eleanor, like these twenty donors who are gonna give a mid-level gift, or being aware there are thirteen recurring monthly donors who might churn or be at risk of lapsing, and let’s do something about it. These are all high-impact, worth-your-while actions. You don’t have to troll through the database, you don’t have to have all those tabs open. Hopefully it’s a big time saver, and that shows why less is more, and using AI for precision — these different types of AI — leads to much better fundraising results.

Ann Fellman: Great, so many awesome questions — everybody was like, “oh my gosh, how, when do I get this?” So we are officially launching this on July first, and it will be a new package or offering. If you’re a current Bloomerang customer, you’d upgrade or have a conversation with our account management team about your specific environment — how many constituents you have in your database — and then the pricing, moving into the Intelligent Giving Platform bundle. So lots of great stuff, there’s gonna be a lot happening throughout the course of July if that’s something you’re interested in.

But while we’re doing that, we’ve gotta do one of the most important things: give away a fifty-dollar Uber Eats gift card. So I’m gonna go ahead and draw a name.

Natalie Nuss — you are the recipient of the fifty dollars! We will send you an email, Natalie, to get your fifty-dollar Uber Eats gift card. Join us on Thursday, we’re gonna do two, two-fifty — maybe you can buy lunch for your whole team, or people can do all kinds of things with an Uber Eats gift card.

So, lots of great questions — I’ll give everybody a second. If you have any follow-up questions, go ahead and log those into the Q&A panel. Some of you have specific Bloomerang packaging questions and other things — I can funnel those off to the team too, so you can get taken care of. But make sure you raise your hand if you’re interested in learning more and want to explore — please do that now with the poll, so we get your hand raised registered.

Tim, thank you — you enlightened a lot of folks in terms of what’s possible, really exciting work. When we think about how pressed our time is, and how to focus on the donors that require that next best action — a lot of people were talking about major gifts and monthly giving, that seemed to be a really big theme for folks today when they think about those next actions. So thank you, thank you.

Thank you, everybody — be back tomorrow, Meena Das is going to join us for AI and equity, so join us tomorrow for that, same time and place. Then make sure you come back Thursday, where Rachel Muir is going to host us with a fabulous roundtable discussion — we’ll have Tim, we’ll have our speakers from yesterday, Loree and Sara will join us, and Meena and Rachel and myself, and we’re gonna have an awesome conversation and address a lot of your questions. So thanks for joining us today, have an awesome day, hope to see you tomorrow and again on Thursday. Have a great day. Thanks, everyone.