Mallory Erickson  1:48  

Welcome, everyone, I am so excited to be here today with my friend Nathan Chapelle, Nathan, welcome to What the Fundraising.

Nathan Chapell  1:55  

Thanks for having me. Mallory, it’s great to be here with you.

Mallory Erickson  1:58  

Why don’t we start with you just telling everyone a little bit about yourself, what brought you to your work in the first place and tell everyone a little bit about what you do today,

Nathan Chapell  2:07  

I spent 20 years in the fundraising world after actually I’ll go back a little bit I was a technology business owner before getting into the nonprofit world spent 20 years in nonprofit. And now I find myself looking in that 20 years working for a private sector company, which is donor search AI, where we only work with nonprofit organizations. And I guess the way I found myself into that was working at a large cancer hospital, realizing that there was technology in the world that could benefit us back in 2017. And just decided to jump into the deep end of the pool and figure out how to use AI. And here we are.

Mallory Erickson  2:46  

So tell everyone a little bit about like what the AI is like that you work with? Because I think there has obviously been a new understanding of AI in the last few months with the rise of chat GPT. But I would say in general, AI still feels a little bit like this gray area that not a lot of folks understand. So can you explain the type of AI you’re interacting with?

Nathan Chapell  3:12  

Yeah, that’s a great question. I have been working in AI before, it was actually cool to talk about, you know, back in the day, I mean, like five years ago, working in the type of AI that I’m specifically working every day is machine learning and deep learning. And some natural language processing. We it’s a AI is a confluence of different technologies that tend to work together. But essentially, my hypothesis in 2017 was, if we could take all of our hospitals internal data that we had access to, we can enrich that data with external data, like consumer data and some experiential data, then we essentially could get away from this idea of trying to essentially find potential donors based on their wealth, which is what most nonprofits still do today. And it’s what we were doing up until this time, instead of looking and prioritizing people based on their wealth, we could prioritize people based on their connection to the organization. So really using machine learning deep learning to measure a person’s connection to an organization, and then be able to do that essentially, in real time over and over again.

Mallory Erickson  4:15  

Okay, so can we talk about this piece of connection? Like, how do you measure connection? Let’s start there, and then I’ll ask some follow up questions.

Nathan Chapell  4:25  

Yeah, I mean, I think if you go from what it was to what it is, now, it’s a pretty stark comparison. In the past, we used to take around 18 to 24 data points to essentially determine if someone was a decent prospect. And those data points would be like, you’re a board member, or you made a gift in the last 12 months, or you graduated from our school. What AI is enabled us to do is do that but on steroids. So an average machine learning model that we use to measure connection uses on average about 1000 data points. And in fact, it’s not just one algorithm And we use on average around 2700 algorithms per client. So lots of algorithms using lots of data. And essentially what it does is it demystifies what we think we know about our donors to actually proving to the 10th decimal point what why donors actually make a gift. And so every organization will look different, because every organization has different demographics and geographics and different culture of philanthropy, and how they foster the low end of the pipeline, or the high end of the pipeline. But they are almost always things that are very intuitive. A good example of this is like, when we’re using lots of data to measure connection, we work with a lot of children’s hospitals. And in some cases, we often find that of all the individuals that will give to a children’s hospital will find instances where they have recently bought children’s apparel in their consumer data file. So this is again, intuitive because out of the world of people that might give to a children’s hospital, people that recently bought children’s apparel, it’s a nice slice off the top to be like, Oh, okay, we don’t have to tell the model. This is a children’s hospital, or even what a children’s hospital is, or what children’s apparel is, it will be able to find the correlations just in the data of how often that data point of children’s apparel shows up in their data.

Mallory Erickson  6:19  

Okay, so in addition to the fact that AI gives us the capability and capacity to analyze all these data points that we could never do alone as humans. What are things that you’ve seen as folks have been using your AI, where you watch fundraisers get surprised by what the connection points show versus some assumptions they’ve been making?

Nathan Chapell  6:45  

Yeah, I think it’s, there’s been a few really big revelations, we call them the evolutions are the five evolutions of AI. And these are things that we really didn’t know in 2017. But we like to say we know more about the true motivations of generosity than ever before in history. Some of the biggest epiphanies have been, some of them are basic, like, you need more than 18 data points, you need to enrich your file, it’d be much more holistic, less bias if you are enriching your data. But two of the biggest data epiphanies that we’ve had is one is almost every nonprofit organization does some sort of modeling, and they try to figure out who their best prospects are, and 99 times out of 100, they will have one model that serves both prospects and donors. And the challenge with this, and I’ve done this for 20 years, is that when you model a prospect against a donor 100% of the time a prospect will score lower than a donor. So mainly because they’re essentially penalize the model penalizes non donors for having not made a gift. But side by side, we know that well, there could be people that are as deeply connected to our mission that just haven’t made a gift. Because philanthropy does not equal money that equals a heart in alignment of values. So when we decided a few years ago, to bite the bullet, and separate donors from non donors and model them separately, we can look at their models side by side and realize how vastly different they are. And so that reduce a tremendous amount of bias that we have toward non donors and donors repeat donors, so that was probably one of our biggest epiphanies. And it’s something that we hope that reverses this bad practice of just trying to fit everyone into that bucket. And the second big epiphany is that most organizations over time would just model someone one time. And maybe one time a year, maybe one time, every three years, that was kind of the customary process. But realizing that people every day have an opportunity to do something or not do something like they can open an email from you or not, or they can volunteer or not, or they can make a gift or not. So the second really big epiphany is that people are not static people are an n of one, they’re an individual that every day has an opportunity in life, to reach out to you or not. And so therefore, to follow up how the private sector measures, Amazon doesn’t just look at me once a year and decides if I’m a good consumer, they look at me every time I open an app and buy something like something in return something, share something, any of that and essentially continuing to reassess my engagement with different things. So those are probably the two biggest epiphanies. And I think those are extremely healthy for organization because, in fact, none of our models, we don’t use wealth data in any of our models, we actually have taken all wealth data out because we proved early on that wealth and giving are very low correlations, we actually just look at connection.

Mallory Erickson  9:43  

I love that I think you answered one of my other questions, which was what is one of the biggest misconceptions around a tool like yours, and I think it is that it’s based on wealth. And so I really appreciate you explaining that and clarifying that to folks and I really love what you said about the fact that our donors are changing constantly. And I’ve been thinking about that a lot with identity giving, I just had a client who lost a donor. And the donor was really straightforward about their shifting priorities. And the fundraiser was obviously disappointed. But when we looked at it, it was like, look like the identity of the stoner has really changed. And what they’re aligned with today is different than what they were aligned with five years ago when they signed up for your monthly giving program. And so I think that’s such an important piece to just testing that engagement consistently, and recognizing that they change over time.

Nathan Chapell  10:40  

Yeah, it’s really allowed us to move away from this idea of personas in generalized personas of like, you’re a donor, you’re not a donor, or you’re this type of donor and really look at this idea of like n of one, which is really, if we follow the guise of medicine, precision medicine is based on this idea that every single person is an n of one, every single person has a different disease based on if you have a type of cancer, you don’t have a general form of cancer, you don’t have prostate cancer, or breast cancer, you have a cancer that’s specific to you, based on your genotype, that every person is treated based on their own DNA. And we’re really trying to bring that level of sophistication to the nonprofit world, that every person is unique to them. And that person can see your point will change over time. You know, not long ago. But actually, when I started, my last nonprofit job was, we haven’t screened our database in three years. And by the way, it wasn’t the whole database, it was only in people that we felt worthy of screening, to the point of measuring every single person, and they’re connected to organization all the time in real time.

Mallory Erickson  11:47  

That is such a good point. And I have really seen that as well, and how much bias we can put into a system like that, and then get the data that we have set ourselves up to get for folks who are maybe new to this idea, or this type of tool. How does using a AI like donor search AI impact the decision making of fundraisers? What types of decisions and actions does it inform for the fundraiser?

Nathan Chapell  12:18  

I mean, it doesn’t matter how great your data is, or how impressive your algorithms are, if you’re not making quality predictions and better decisions, right. And so, first and foremost, I think the types of decisions that you’re making, one are more donor centric and less bias in the sense that they’re prioritizing rich people that are likely to give, which is unfortunate, I laugh, but literally happening nine out of 10 times in our industry today. So when we see and work with, I think there’s a couple of big changes is one is that AI is very different in the sense that AI, you’ll never be ready to do it, and you’ll never be done with it. So it’s there’s freedom in knowing that when you start machine learning, the operative word is learning or deep learning is learning and it’s getting better over time. What that means is that every person who receives the information is part of the feedback loop. So fundraisers aren’t now just the recipients of data. They are now part of the feedback loop. And so what I see is a big shift and a culture shift where we create this equilibrium of art and science is that now it’s not just like I’m in charge of the data and I pass out prospects like I pass out candy. It’s like my job is to make the candy and give it to you. Now it’s our job is to take all the data, we have a bowl of candy to enable fundraisers to take the candy that’s best for them. It’s really like you think about a playlist and the difference between curating your own playlist versus curating playlists of songs you like and then having a Spotify, augment that and enrich it with some other songs you might like not ultimately be perfect, but surprisingly, it actually gets better and better over time. It’s very similar. And I think that’s a process and the change management that our industry is going through right now and will continue to go through because the haves and the have nots of AI and non AI are going to be very noticeable near future and things like Chachi Beatty just said they put a spotlight on that aspects and how quickly AI can elevate your role versus someone that is not using it.

Mallory Erickson  14:23  

Yeah, I so appreciate everything you said. What do you think are fundraisers biggest fears when it comes to AI like when you all have recommended AI in forming certain decision making? Or what is that fear that comes up?

Nathan Chapell  14:37  

What part of the fear is grounded? Because if we think about AI in the private sector, I mean there’s been lots of horror stories right I mean, Twitter, of agents, racist Islamophobic algorithm ablest and all kinds of bias and people hear that they remember that and Twitter stock price went down. The fear for most individuals is that there’s this differential that philanthropy is entirely based on trust. And trust is the currency of the nonprofit sector. How do we trust something, if there’s a black box, we just call it the black box. I don’t know what’s in it. I don’t know how it’s making these algorithms. And we made a commitment early on that. And I’m a very big proponent of this personally, that there should be no such thing as a black box in the nonprofit sector. So any model we build can be accessed viewed visually by anyone. And so I think one of the biggest fears is the fear of like, I don’t know what’s in it. But once you understand what’s in it, and you can see it visually, and you can see how it’s making its predictions. There’s no hypothesis, it’s just math. And at the end of the day, if you have more people that are making their first gift, because they open alumni newsletter on a Tuesday or Thursday between two and 4pm, then that is a strategic data point you can use to do more of those things. So I think it’s more of a fear of the unknown. And I think our industry has created some of that fear, because we have created black boxes, in some cases, and in some of those black boxes, and early AI tools, made giving more transactional. But I think anytime you were talking about AI in the nonprofit sector has had the word responsible ahead of it, that there should be like responsible AI, there’s no AI in the nonprofit sector, that should not have the word responsible in front of it. And that’s something that we have to hold each other to a high level of accountability to make sure that we are competitors, or not anyone doing this needs to maintain privacy and security, and ethics and equity among all the data that they’re using.

Mallory Erickson  16:38  

Yeah, what is the problem that AI solves. And you can share a little bit about generosity crisis, if you want in this too, like, I’m just thinking about the role, you see AI playing in solving some of the biggest challenges that we’re facing as a sector.

Nathan Chapell  16:56  

I firmly believe that AI is the only scalable solution that can help reverse the generosity crisis. You know, of course, in America, the percentage of households I give to charity has been decreasing at a pretty increased rate. And that doesn’t end well for anyone, I mean, giving ends in 49 years to traditional nonprofits, that something doesn’t change. But when we spent so much time researching and looking at our book of like, telling people to go back to church and just go to church, and we’ll solve the generosity crisis, it’s not something that is going to go over very well, AI, because true AI is evaluated on precision and personalization. If you can leverage precision and personalization at scale, affordably, and then essentially let that serve as this like Copilot to your prospect development efforts, we can increase the things that are needed so badly in our field, because people the 51% of Americans that don’t give to charity, they don’t give because a it’s either not personalized, or they’re not the right fit for your organization. So it’s led a lot of behavior, like spray and pray and lots of other things where it’s just worn people out to the point that most non donors feel like, you don’t really know me, you don’t really care about me, I’m just an ATM to you. Therefore, I’m not going to give. I think AI has the ability to solve for that and super bias in this. But I’ve kind of bet my entire career on the fact that is the only scalable solution that can help reverse it.

Mallory Erickson  18:23  

I really appreciate you sharing that. And I think as somebody who has been tech timid throughout my life, I think it’s taken me some time to really sink into the truth of that. And to really deal with my own dis. I love what you said before about the black box, I think so much of my discomfort with AI or automations even was like, I don’t understand how this works. And that makes me nervous to use it. And it’s taken me a lot of time to just be really honest with myself about like, what these tools open up. And how can I be comfortable and okay with not always totally understanding it, like my husband could talk to you about these models till you’re blue in the face and I sit down, I’m like, can you tell me a picture? And I think like That’s okay, like, I want people to feel like they can embrace AI and overcome the discomfort that comes with not understanding it, instead of feeling like they have to totally get it to use it.

Nathan Chapell  19:25  

Yeah, what scares me is people that run algorithms that can’t explain it themselves. So there’s grounded fear in that where you as an individual to look at like 3000 lines of code, it’s not going to make you feel more comfortable. But if you can see your algorithm visually, and to look at data points that are yielding the greatest part of the prediction, it starts to make sense. Now there has to be that willingness that responsible AI starts with this willingness of an organization like ours to say like we will only build models for you that are fully transparent, that you have full access to and that you’re part of because If you are part of that feedback loop, we are just maintaining and helping retrain. So when you realize that AI can remove tremendous amounts of bias, and the biggest was very early on in my AI career, was realizing that we were adding so much bias by our models, if I was to characterize you with 18 data points, which is the average number of data points in a classic regression model, how much would we be missing a view? Who am I to say, well, these are the most important 18 data points. So Mallory versus Let’s use 1000 data points and let the model actually do its thing and help us. And so still, you need software and people to essentially look actively for bias. And that has to be part of the responsible AI. But I didn’t realize how much bias we were adding into our assumptions of which 18 data points were important to begin with. And we’ve challenged a lot of that, which is freeing, actually. And once you cross that calm, then you feel like, wow, this is actually a realize like I’m reducing bias. I’m more holistic, I’m conscious of this, I know what’s in it. And actually, a lot of the data inside an algorithm is very actionable. from a strategy perspective.

Mallory Erickson  21:07  

That’s such good advice. And I think it helps us break out of scarcity mindset. And a lot of ways a lot of times where we have those limiting beliefs around what is and what is impossible with our donor base. I feel like these tools help us see the things we can’t see ourselves. And so I really appreciate that. Any final things you want to make sure to leave folks with I’ll make sure generosity crisis, we have links to the book below. But where should folks go to find you. And any final words,

Nathan Chapell  21:34  

my final word would be, it all starts with starting, I share this and it’s not flippant, but it’s like, you’ll never be ready to use AI. So don’t wait for your data to be perfect. Don’t wait to have more data, you’ll never be ready. In fact, AI is really good at filling in holes and gaps in data. You’ll never be ready and you’ll never be done. It was one of the most freeing moments that I realized that I was conditioned in my career to like build a model, use a model until it wears out and then build another model. It’s just this like horrible kind of feedback loop of just like going through over and over again, feeling like you’re rinse and repeat. Realizing you’ll never be ready and you’ll never be done is actually very freeing that you have a model that you can use to learn and grow and that will get better over time is a really big flip to switch that I think took me at least a year when I was first learning how this works. And then as far as getting connected with me, LinkedIn is really a great way you can also go to the generosity crisis.com To learn more about the book and learn about me and my co authors. Yeah, love LinkedIn. I’m on all like you. Both LinkedIn nerds love our LinkedIn communities. And yeah, happy to connect with anyone that way.

Mallory Erickson  22:42  

Amazing. Thank you so much, Nathan for this conversation and for joining me on the show.

Nathan Chapell  22:48  

Awesome. Thanks, Mallory. It’s such a pleasure to be here.

Mallory Erickson  22:56  

All right, there is so much inside this episode, but here are a few of my top takeaways. Number one, AI technology gives us the ability to uncover the secrets behind personalization and precision in charitable giving that we could not learn another way. Number two, we need as a sector to embrace the discomfort associated with AI implementation if we want to see improved fundraising results. Number three, in richer data file with external data and experiential data to better understand donor connections. Number four separate donors and non donors when modeling to reduce bias and improve understanding of motivations for giving. I thought this was so interesting. Number five, continually reassessed donor engagement, as their preferences and circumstances may change over time, looking at data and running AI models is not a one time thing. Okay. For additional takeaways and tips inside this episode, head on over to Mallory erickson.com backslash podcast to grab the full show notes and resources now. You’ll also find more information there about Nathan generosity crisis and donor search. Thank you for spending this time with us today. If you enjoyed this episode, we would love it if you would give it a rating and review and share it with a friend. I’m so grateful for all of my listeners and the good hard work you’re doing to make our world a better place. And if you miss me between episodes, stop by and say hello on Instagram, under what the fundraising underscore. Have a great day and I’ll see you next week.