Community IT Innovators Nonprofit Technology Topics

Nonprofit Better Deal for Data Pledge with Jim Fruchterman

• Community IT Innovators • Season 7 • Episode 49

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 36:18

Carolyn Woodard explores responsible data governance and AI realism with Jim Fruchterman, MacArthur Fellow and founder of Tech Matters, a tech-for-good nonprofit building open source software for the social sector. Jim's work sits at the intersection of two urgent questions every nonprofit is wrestling with right now: what do you owe the people whose data you collect, and how do you make smart decisions about AI without getting swept up in the hype?

Jim introduces the Better Deal for Data, a new data governance movement built around seven plain-language commitments nonprofits can make to the communities they serve. The core idea: you don't own the data of the people you serve, tech vendors may be extracting it right now without your knowledge, and a basic data safeguarding policy should be as standard as a child safeguarding policy. He also explains why you can't build a responsible AI governance policy without first getting clear on your data governance.

Then the conversation shifts to AI strategy, where Jim draws on decades of experience as an early AI entrepreneur to offer a genuinely grounded take. Gen AI fails in nonprofit program delivery about 80% of the time, and that's still a much better track record than blockchain or the metaverse. 

Jim and Carolyn discuss:

  • The seven Better Deal for Data commitments and why the average nonprofit can likely adopt them in two hours
  • How tech vendors quietly extract and monetize constituent data through survey tools, donor management platforms, and more
  • Why data governance and AI governance are inseparable, and why feeding confidential client data into a free AI tool violates both
  • The case for nonprofits pooling anonymized data to build better AI models for social impact, with real-world examples from MomConnect and Community Solutions
  • Why Jim recommends most nonprofits wait for proven AI products rather than build, and what RAG-based tools are actually delivering results right now
  • Why being two or three years behind the for-profit AI curve might actually put nonprofits five to ten years ahead of where they were last year

Resources Mentioned:

_______________________________
Start a conversation :)

Thanks for listening. 


Carolyn Woodard

Welcome everyone to the Community IT Innovators Technology Topics podcast. I'm Carolyn Woodard, your host, and I'm really happy today to welcome a new guest on the podcast, Jim Fruchterman. So, Jim, would you like to introduce yourself?

Jim Fruchterman

Sure, Carolyn. So I'm a uh a tech nerd, um, a serial entrepreneur from Silicon Valley. Um, I used to start the for-profit kinds of tech companies. Um, I started seven in a 12-year period, and only five failed. Um, and and the two that succeeded were were both what we now call AI companies. Back then we called them pattern recognition or machine learning. And uh, and

Jim Fruchterman

You know, we we raised a lot of money uh for this for this startup from venture capitalists, and uh I created a new product that was going to help the blind and showed it to our board, and they vetoed it on the spot because, well, the market for reading machines for the blind was a million dollars a year, and they'd put $25 million in the company and they just didn't see the connection.

Jim Fruchterman

So, so ever since that board meeting, I've been trying to make the connection between tech and social impact because there's an awful lot of great ideas that are possible with tech, but they're not sufficiently profitable to make venture capitalists excited.

Jim Fruchterman

So, for the last few decades, uh every two or three years, I start a new tech for good company inside a charity to build the product that Silicon Valley won't bother to build because they can't make enough money off of blind people or people in Zambia or human rights activists or, you know, I don't know, 90% of humanity and the planet probably don't make enough money for the average tech company, the way our systems are set up.

Carolyn Woodard

Exactly. Um, and I think uh you and I met when we were at the uh Good Tech Summit in DC a couple of months ago. Um and you talked a lot at that summit about some of your new things that you're working on. So I wanted to ask you like, what are you working on now?

Jim Fruchterman

Well, I mean, I mean, traditionally I I do tech products, right? And so, so, you know, uh, and I'm still connected to all these tech products.

Jim Fruchterman

So probably the best known one I've I've done is Bookshare, which is the National Digital Library for Disabled Kids. So if you're dyslexic or blind, you need a book for school, we'll make it into an ebook that's accessible. Um, our our

Jim Fruchterman

Our two biggest projects right now are Aselo, which is like uh the contact center platform open source, uh kind of like 911 for kids in 20 countries around the world, including the National Child Abuse Hotline here in the US. Um uh

Jim Fruchterman

Tarasso, uh, like what if people uh on the front lines of climate change actually had someone building software for them? So what uh we have the an open source soil identification tool. Uh,

Jim Fruchterman

We have an open source version of story mapping, which is how do you take map data and other data and pictures and videos and make a compelling story about, I don't know, why you should visit someplace that has ecotourism or uh stop doing something terrible or invest in them, whatever it might be. Uh, but

Jim Fruchterman

But in the last few years, I've been spending a greater and greater percentage of my time on just advancing technology for good generally. And and so that shows up in the form of a podcast, of course, you know, where I interview great tech for good leaders, following the footsteps of some people who are been doing it longer and uh for than I have. Um uh

Jim Fruchterman

I just wrote a book called Technology for Good, you know, which is um the first book on how to start a tech company and not make money. So it's aimed at the person who wants to start a nonprofit tech enterprise, either inside their nonprofit or or you know, uh from scratch as a software company or a hardware company. Um, and so and

Jim Fruchterman

I talk about 60 different tech for good nonprofits who've put technology at the core of their program delivery. In other words, the you know, they they have some innovation, but the technology makes it possible to go to scale. Um, and of course, I also talk about the many examples of failure in our system, because you know, that's the nature of tech. Um, and uh, and

Jim Fruchterman

Earlier this year, we launched a responsible data governance initiative called the Better Deal for Data. Uh, it one of the things that people were asking us about is so, you know, what what are the promises that nonprofits make to vulnerable communities about the data they collect? And the answers are generally none. Um, and you know, we work with kids. No one who works with kids can get funded by a responsible foundation unless you have a child safeguarding policy.

Jim Fruchterman

Why on earth are we working with vulnerable people as a sector and no one thinks about a data safeguarding policy or constraining the data use? You know, maybe don't sell the data of poor people to the fintech industry so they can, I don't know, squeeze them harder. So

Jim Fruchterman

So the idea of the Better Deal for Data is just seven one-sentence commitments that I think would make sense to any nonprofit leader, you know, and these are all they're in the form of promises the nonprofit makes to their stakeholders, who who, you know, obviously the people we serve, whether we call them beneficiaries or users or stakeholders, um, donors, the employees. I mean, we're handling all this data. Well,

Jim Fruchterman

First, uh, the promise is it's your data, it's not ours. We don't claim ownership in it. If you ask us to delete it or correct it or transfer it to you, we will. Um, if uh if we do research on your data, uh, we'll anonymize your data and you get a free copy of the research uh rather than having to pay for for that. Um, we're primarily collecting this for social good reasons. In other words, that's our primary purpose. We're here to help you and your community, maybe science, humanity, and the planet, but we're not here to help private interests. We're not here to make a buck. We will not sell your data to the meta 250,000 companies that are all spying on us. Um and uh anyone who touches your data will make a legally binding promise to the above.

Carolyn Woodard

Interesting.

Jim Fruchterman

Yeah. And so, so, you know, I think the average leader would go, no, it seems reasonable. I mean, we're a nonprofit. Why, why on earth would we sell out the people we serve? But

Jim Fruchterman

But the tech companies, unbeknownst to a lot of a lot of nonprofits, are stealing the data of our stakeholders right under our nose. I mean, there are survey tools that copy everything that anyone enters into your survey. Uh, you know, there are donor management tools where you know you you take a donation from someone and they say, oh, Mary Smith, let's add her. Oh, we already know about Mary Smith. Let's now you know enhance our database and sell this to other people. I don't think,

Jim Fruchterman

I think the whole thing is people expect that nonprofits will collect data about them. People expect that nonprofits raise money by citing their their impacts. People do not expect to be so to be sold out. And so so it's a lot of it's around how nonprofits can be as trustworthy, as trust trustworthy as nonprofits should be, and uh how to not surprise people. I mean, our our idea is if we tell you what you're doing, what we're doing with your data, you should go, oh yeah, that makes sense, as opposed to .. "you're what?!"

Carolyn Woodard

And then uh it seems like this is such an like you said, an obvious set of principles to sign on to and to help uh your develop your governance of that data and those databases. Do you have you know additional supports for nonprofits who may be struggling to have security around their database or with which vendors they're using, as you said, that will be spying on the data or not spying on the data?

Jim Fruchterman

Well, I think I think our our goal, I mean, this is brand new, right? So we're in in tech terms, we're beta testing this, right? The you know, the early adopters are are asking us questions. I mean, I mean, people are saying, hey, it says you're not supposed to monetize the data, but of course I raise money based on my data. I'm like, well, based on your aggregate data, right? I mean, I I think

Jim Fruchterman

I think the fact that you served, you know, 4,700 people last year is not sensitive. But just don't name them all by name, right? You know, that would be bad, right?

Jim Fruchterman

And so so we're busy working through these explanations. Um uh people are worried that it's complicated. Um, we were

Jim Fruchterman

I'm trying to come up with a adopting the Better Deal for Data in two hours uh little thing. Like, you know, and and and because the average community-based organization is not doing terrible things with your data. And we don't care if you use Google Analytics. We're not being that kind of picky.

Jim Fruchterman

We're just saying, you know, the the I mean, you're you're serving kids. Well, you should safeguard the data of kids. You're serving domestic violence survivors, you're you're you're you're serving immigrants, right? Maybe be really careful. Maybe not even don't collect their immigration status, because that's the best way to protect them, is just not to have the sensitive data. So, so we're in that sort of zone.

Jim Fruchterman

If if it takes off, which I hope it does, then we're gonna do more about um like certification. You know, people say, Oh, can you certify us? Like, look, we're one and a half people working on this. Let's let's let's not get ahead of us on this. So, so but I think that, you know, the next year we'll get kind of a feeling for this and then decide, you know, are are we helping people adopt this?

Jim Fruchterman

Are we gonna, like a lot of consultants exist to help you comply with standards, are we gonna be supporting a bunch of consultants? But but my dream is that the average community-based organization who's never talked to a lawyer can look at these things and say, yeah, we can agree to that. We'll put a we'll put it on our website that we we don't sell your data, that we that we comply with the better deal for data. And I want to make that possible.

Jim Fruchterman

Whereas someone who's, I don't know, a giant international NGO that's collecting 15 different kinds of data, if you know, if they sign up for this, they're gonna have to actually like think about all the things they're doing with data. That's a longer process. Um

Carolyn Woodard

But they have lawyers.

Jim Fruchterman

Yeah, they do have lawyers, yeah, policy people. Um, but my my uh we're also uh suggesting that people endorse the standard. Uh-huh. You know, this is a good idea. You know, maybe we'll adopt it later. Um,

Jim Fruchterman

But I think it's a it's a signal around, you know, rejecting the surveillance capitalism sort of system that we have. Um, and yes,

Jim Fruchterman

Over time, I do want to put pressure on tech vendors. Uh, let's say, for example, we work with child helplines, right? That's one of our major areas. I mean, I want the International Association to say that all child helplines should adopt this policy as part of their professional obligations. And then anyone who wants to sell to the child helpline movement has to not steal their data.

Jim Fruchterman

And I think that individual nonprofits have a hard time negotiating with big tech. But I think that our sector and and our donors have more leverage if we act together. Um, and, you know,

Jim Fruchterman

I mean, if you want to use a survey tool, there are plenty of survey tools that don't steal your data and and sell to you based on the fact that they won't steal your data. You should use them as opposed to the people who are mysteriously silent and just refer vaguely to a 75-page terms of service that on, you know, page 67 claim all your data and your firstborn.

Carolyn Woodard

Right, right, right.

Jim Fruchterman

Because they can, you know.

Carolyn Woodard

So I love that this seems to be tapping into this like bigger question about the ownership of the data and your right to own your own data, which I think has been percolating in Europe for a while. It seems like it's gaining some ground here of that data ownership being a justice issue. And one that foundations and philanthropy have kind of quietly like extracted that the insights from that data, maybe in ways that weren't transparent to the populations or to the the communities that they were working with. So it feels like a good foundation to and a good moment to be to be starting this pledge.

Jim Fruchterman

Yeah. And and we have, I mean, let's pick indigenous communities around the world. Yeah, they've been on the short end of colonial practices for a little while. And um, and they have really good data governance standards. So, for example, you know, our dream of the Better Deal for Data is not replacing indigenous data governance. They've got that down.

Jim Fruchterman

It's to say to the nonprofits that work with them, honor their data governance structure by not, but if you don't steal their data, then you're probably honoring their existing governance structures.

Jim Fruchterman

But yes, uh black and brown communities are worried about how data is used against them because it's been used against them for a long time. Um, and so so this is this is kind of familiar territory. And then, you know,

Jim Fruchterman

I just came back from yesterday from Africa, and you know, more and more countries are saying we don't want to have all the data on our citizens being held by big American tech companies. And they've got a pretty good basis to be skeptical of why this is good for them.

Carolyn Woodard

I think that was going to be my next question is we've seen a lot of our clients and um, you know, nonprofits coming back to think about data, data policies, data retention, privacy policies because of the advent of AI and AI search and the difference that, you know, maybe this is something that, oh, we probably should do that, you know, for like a decade or so. We're like, oh, we should clean up our data, we should know what our data is, we should have consent of everyone that's in our database. And now it's become really urgent.

Carolyn Woodard

So could you talk a little bit about like this new, like how AI interacts with this data question?

Jim Fruchterman

Yeah, I think I think you're you're right in identifying that data should have been something that we are dealing with. Nonprofits should be good with their data, right? Nonprofits should not be using data in bad ways.

Jim Fruchterman

But because AI basically feeds on data, and it's trying, you know, it has this gigantic appetite for data, a lot of the value you get out of AI relies on feeding it data. So now that I mean, so people are often talking about let what's our AI governance policy? Um, but you know,

Jim Fruchterman

I think that you can't do an AI governance policy without having a data governance policy. They they kind of go hand in hand.

Jim Fruchterman

And of course, the Better Deal for Data, you know, if you know, we have the seven, you know, one-sentence commitments, but there's a few thousand more words about what this means and doesn't mean.

Jim Fruchterman

For example, one of the things that you can't comply with the better deal for data and feed confidential client data into a free version of open AI where they're going to keep a copy of your client data and train on it and make commercial money off of it. That's not okay. Um, and so so

Jim Fruchterman

I think this is so I think people are more concerned about AI governance because of their worries about AI. But I think I think we need to be working on both of these tracks of, well, let's don't do data and AI governance, and let's keep, let's keep our eye on the idea that we're going to be doing better with both. And and so uh so

Jim Fruchterman

One of the dreams that I have is that um, you know, we're we're a data organization, we've worked with a lot of sophisticated nonprofits, we've negotiated data sharing agreements, and you know, that takes lawyers and takes a year or two, right, to do these things. Um, my hope is that if you have, let's take, you know, 10 agricultural co-ops or 10 nonprofits working in housing in a metro area, that that if they've all adopted something like the Better Deal for Data, they could pool their data for social purposes, you know, and use it to say, ah, you know, now that we have enough data, we can actually build a better AI model of climate smart agriculture. Or um, what are the leading indicators of someone who's going to be successful in transitioning out of temporary housing into long-term housing?

Jim Fruchterman

Because, I mean, the the challenge around AI is that, you know, whatever it is, five big companies have data on a billion people and they're part of an ecosystem that is just you know constantly vacuum cleaning up all of this data.

Jim Fruchterman

The nonprofit sector, all the data is like in a million tiny silos. And and so how do we actually unleash that, not for profit, but for social impact, so we can build better AI models? Um, and

Jim Fruchterman

I mean, many nonprofits would love to know how they stack up against their peers, not what specific peers are doing, but benchmarking themselves. These are all things that we could do if we had, I don't know, a rational, basic set of data governance sort of structures that say, yeah, yeah, doing AI for good, sure, as long as it respects the confidentiality of the people themselves. You know, doing, you know, combining data sets so you can get a better handle on what's going on in your metro area, sure. Or whatever it might be.

Jim Fruchterman

These are the sort of things that business is already doing in search of the buck. Why aren't we doing this in search of more social impact?

Carolyn Woodard

Do you have any example? This seems like something that, say, a foundation has been working in, like, you know, education. So they have grantees that have a lot of data. Um, are they are there situations where they have done this kind of aggregating the data from these different silos that could be a model? Or do you imagine it's like, you know, a virtual database with security all around it, and you like put your data into the data bank and then you get to ask it questions about the other data that's in there?

Jim Fruchterman

Oh, yeah, we have tons of models doing this sort of thing across the social goods sector.

Jim Fruchterman

But I would say um, if we go further up, uh science has been doing this pretty well for years, right? You know, if I'm a young astronomer, I don't have to get time assigned on the space telescope. I have the last 10 years of space telescope data available for me to mine and and try out on data that's already been collected, right? So, so uh and and

Jim Fruchterman

In the social goods sector, you know, the gold standard is in medicine and public health, right? We have a long tradition of collecting data, measuring impact, and safeguarding the identity of individuals in the health system. And in the US, you know, this is called HIPAA, right? Which is our medical privacy law. Um, and it's you know, it's an older law, it's got its challenges, but fundamentally the social bargain is, you know, so I I disclose something very confidential and sensitive. I get help from a doctor, um, it's kept secret, and my data gets used in medical research. And by and large, people are happy with that social bargain. They think that's a fair bargain.

Jim Fruchterman

Well, how do we do that in sort of the nonprofit sector? So I'll I'll pick an example of um so uh academics in in some areas are fiercely protective of their data. Well, the the Gates Foundation, to pick an example, a big example, said, you know, hey, if we're paying for this data collection, um, you're gonna make that available, that data available, not not openly if it's confidential data, but you know, uh if we want to do a bigger study. And so

Jim Fruchterman

I remember talking to someone from Gates not too long ago, um, where they had actually, they'd paid, you know, Gates, huge, they'd paid for a whole bunch of studies about stunting, which is where when kids are malnourished, um, you know, they they don't grow as tall, they don't grow as smart because as a kid, it's a very sensitive time to be starving when you're young.

Jim Fruchterman

But one of the questions was when is the critical period to to intervene? Um is it any time in zero to three, or is zero to six or nine months more critical than zero to three? And it turns out by doing this big study across a bunch of their grantees, that zero to six months turns out to be really crucial. You know, you you want to make sure to intervene there. Certainly, you know, starving when you're a two-year-old is bad, but it's not as bad as starving as an infant. And so so

Jim Fruchterman

That was something where the individual studies couldn't have picked that up, but by by combining them. And you know, something that a lot of people don't know is there are gigantic medical databases held at our major research universities under pretty tight control, but you can access, you know, brain scans and DNA of uh individuals under this, you know, in this very tightly controlled sandbox where you don't get to figure out who had brain cancer. You just get to figure out uh whether your research thesis works or not and get the answer. Because you don't need the individual's data, but you do need the raw data to actually like I don't know. It's it's kind of it people figure this out. We're trying to actually spread this.

Jim Fruchterman

So I'll give you a last example, and this is something you should be kind of familiar with, some of the battles between different kinds of vendors. And I'll I'll simplify saying it's Apple versus Google. You know, Google's tracking your every move, right? And if you have an Android phone, it's very hard to turn off that tracking everywhere. Um

Jim Fruchterman

Apple uh doesn't actually track your individual movements. What they do is they they send a scramble, a scrambled version of your location data. So they can they can do the same kind of analysis of saying what are commute times, but they aren't tracking Joe average and exactly what path he's taking. Because they're doing the AI and the scrambled data.

Jim Fruchterman

And this is called federated learning. Google also does it in many, many areas, but the idea that we can get the benefit of mass data collection without actually having a detailed dossier of your data. And so it takes more data to do AI that way, but a lot of people are a lot more comfortable. And you know, and Apple is, you know, selling you on that as opposed to their competitors.

Carolyn Woodard

Oh, that makes sense. And I just um so I would be so interested to see what happens with this idea of thank you for explaining it so clearly, of taking that kind of medical or you know, research-based database, but actually making it accessible to nonprofits data to be used in that way, anonymous but valuable.

Jim Fruchterman

Yeah, or or under very tight, you know, legal controls. That's I mean, that's another way you can solve this. I mean, uh, you know, uh uh

Jim Fruchterman

Community Solutions, who famously won the MacArthur 100 million and change challenge, they're working on homelessness. Um, and you know, one of the things they do is that rather than measuring beds or meals, you know, served, kind of measuring the the outputs, um, they're trying to say, is everyone who was unhoused in our community three months ago, are they housed now? Well, that requires tracking individuals through the system, through a whole bunch of organizations, because you know, a homeless person might touch 10 different agencies in that three month period.

Jim Fruchterman

And so, so they have to, you know, constrain what the data is used for, but actually working on reducing homelessness to functional zero is what they call it, is a much better way of doing it than just measuring the outputs of the homelessness industrial complex.

Carolyn Woodard

Right, right. I

Carolyn Woodard

I want to pivot just a bit because I know that you are known as a bit of an AI skeptic. So I wanted to get your your take on uh kind of AI hype and nonprofits in this moment, because you know, we're seeing and hearing a lot of like there's some really big grants coming out for like big projects to like totally transform your nonprofit doing this and that with um you know from Google or what have you. Um

Carolyn Woodard

So what is your advice or what is your take on kind of AI and nonprofits right now?

Jim Fruchterman

I realize I started AI companies in Silicon Valley in the 1980s, right? As a very, very young person. And and I've been doing AI for social impact ever since. So I am an AI enthusiast.

Jim Fruchterman

What I am not enthusiastic about is AI moronic hype, you know, you know, and and you know, in in my book, I I use the example of a you know, the National Eating Disorders Association fired all their human counselors and replaced it with, you know, a chatbot, right? Tessa, the AI-powered chatbot, who was found to tell someone to count calories within three days of launch, and they had to shut it down because counting calories is the last thing you tell someone who's got an eating disorder.

Jim Fruchterman

So, so people have bought the hype and then just cratered big time. And so, um, so the the thing though is, and I'll I'll contrast this. You know, I I say 80 or 90 percent of AI projects in program. Again, I'm talking about program, the most sensitive part of a nonprofit is program. I don't care if you're using it to write a better grant application. Fine. Spell checker on steroids, go for it, right? You know, you know, I'm I because you're not betraying confidential data unless you don't mind them training on your grant proposals, but I don't know, probably probably not the biggest thing to worry about. But um,

Jim Fruchterman

But in program, you know, I say, you know, 80 or 90 percent of the time it fails. And people say, Oh, you're you're a skeptic. And I'm like, no, I mean, blockchain failed 99.99% of the time, metaverse failed 99.99% of the time. A Gen AI failing 80% of the time, that's a 20% success rate. That's huge. That's great.

Jim Fruchterman

And so my job is to help people to, you know, because it's so common for people in the nonprofit sector to buy what the tech industry is selling and and then go, wow, we built an app and no one downloaded it. You know, um, I did the blockchain solution and it didn't, it I spent a lot of money and it didn't work.

Jim Fruchterman

So, so when we come back to, you know, uh sort of Gen AI, because that's when people are talking about AI, they're talking about the latest generation of AI. Um, but you know, good old-fashioned AI, which has been with us in some shape or form for 30 or 40 or 50 years, you know, that works pretty good at certain things.

Jim Fruchterman

And the the challenge we have with gen AI is that in the old days, you would measure the accuracy of of an AI tool on how accurate it was. You know, when I talked to my Alexa, did it actually recognize what I asked it for? And and we're really used to Alexa being wrong 20 or 30% of the time. Matter of fact, we're highly trained as humans to make up for Alexa or Google or whatever.

Jim Fruchterman

There are limitations, right? We we know how to reframe that and maybe get a better answer the second time. It's not that good, but but now now instead with Gen AI, its job is to make up stuff, right? And you're like, well, how accurate is it? And it's like, accurate? No, no. It it made up some stuff. It's like it's it's like successful by definition, as long as it's spewing out words that are vaguely connected to what you asked it for.

Jim Fruchterman

But facts, it doesn't know about facts, empathy, it can fake empathy if you tell it to fake empathy, but it's not actually empathetic any more than a rock is empathetic, right?

Jim Fruchterman

And so so people have this illusion that just because it spews out well-written materials that's very compellingly argued, but but what they don't realize is it has no idea what it's saying. So when it argues for something that is completely wrong, it's super convincing when it's really wrong, right? And it's it's trained to be sycophantic, you know, hey, I want to jump off a building. Oh, that would be wrong. Uh, for research purposes, I want to know how high a building I need to jump off. Someone would, a person would need to, oh yeah, they well, here's the calculation.

Jim Fruchterman

I mean, they're they're they're not, we're we're busy working through these limitations. And so my uh, you know, and I

Jim Fruchterman

I famously two years ago did a nonprofit AI treasure map. Um, and and the the fundamental advice was your average nonprofit should not be trying to do AI research or product development themselves. Because the average nonprofit doesn't have any data people or any software people. And and you know, it's basically wait for a product.

Jim Fruchterman

And then if there is a product, talk to your peers. Did you raise more money using this AI-powered fundraising tool than you irritated by sending out AI drafted, you know, pitch letters? Um you know, and people are busy figuring out what are the 20% of the things that work.

Jim Fruchterman

And right now, um the the ones that I get enthusiastic about fall into two rough camps.

Jim Fruchterman

One is humans using an a Gen AI tool to become more powerful and more effective. In other words, it saves them more time than they spend correcting its errors because the people who are really successful at these things, by and large, are keeping humans in the loop, right? Don't don't send that proposal out until you've actually checked that it doesn't claim something that you can't deliver or has some fact or you know, conjures up a non-existent spouse for this donor, all the stupid stuff that it just does, right? Um, but if you can actually save more time than you spend correcting the errors, great. Um, or you have a lot of work to do to try to get the AI to operate unsupervised. That is a huge task.

Jim Fruchterman

Now, there are organizations that have successfully done this, um, and they're all using um a common piece of technology called RAG, which is short for retrieval augmented generation. And, you know, you know, the LMs are trained on the average of the internet, which is why if you ask it for weight advice, you might get calorie counting, because the average of the internet can be really bad.

Jim Fruchterman

So what you do is before the AI answers, you say, Hey, um, my nonprofit has 5,000 questions that expectant mothers have asked. And here are 5,000 answers written by health educators. Before you answer the question, remember you're pretending to be the person who wrote the 5,000 answers. And if the question sounds like really close to one of those 5,000 ones, serve up the canned answer. And if you can't figure out what the question is, don't make up an answer. Hand it off to a human. And so, so

Jim Fruchterman

Mom Connect, which is the number one maternal advice, you know, hotline in South Africa. About half of their questions get answered by the chatbot based on it being very close to a question someone has already asked and been answered, and it's available in all eight of their major language groups. And half of the questions go to a human because the AI didn't understand it, or this woman is asking three questions simultaneously, and a human needs to have the judgment of saying, that third one is the one that's going to send you to the clinic right now. We're not even going to talk about the first two. Go, you know, or whatever it might be.

Jim Fruchterman

The AI is not that smart to be able to do that kind of thing at this point. So when I talk to my friends in Silicon Valley, because that's where I'm based, you know, everyone says, Oh, RAG, that's so three years ago. You should be using agents.

Jim Fruchterman

And I'm like, no, nonprofits should not be running off and going off in the agent world. Why? Because the for-profit world hasn't figured out how to make the agent AI thing work, even though they're all selling it like crazy.

Jim Fruchterman

I just say, you know, let them spend a few trillion dollars over the next couple of years, and then we'll find out what agentic AI is actually good for. And then the nonprofit sector can be two or three years behind the times, not waste billions of dollars, you know. We'll just um, and uh, and you know,

Jim Fruchterman

Frankly, if we can get the nonprofit sector to be two or three years behind, we've probably just moved them forward five or ten years, you know. So, so you know, it's a we and

Jim Fruchterman

I guess my last point is you know, the industry is there to make gigantic money. And there's gigantic money to be made. I mean, there are people who are getting value out of the personal productivity of this. But if you're running a nonprofit, you know, you're you're not about profits, you're about social impact. You have, you know, a do-no-harm. sort of, you know,

Jim Fruchterman

The for-profit companies, harm, screw it, we're making money. Harm's okay. We're gonna we'll we'll outsource the harm to the nonprofit sector to fix, right?

Jim Fruchterman

But but but as a nonprofit, we have a higher, a higher, I don't know, moral and ethical sort of standard, which is we're here first on behalf of the people we serve. And that means that, you know, we have to be really careful that we're not unleashing bad AI on good people who need help and and are not getting it from the AI.

Carolyn Woodard

That's um a lovely place to land on, I think. I I that's a that's a great summing up of nonprofits and AI. Thank you.

Jim Fruchterman

Oh yeah. Well, and and thank you, Carolyn.

Jim Fruchterman

And and you know, the other thing I'd like to mention is that my book goes open source in September. So so yeah, I mean, it's obviously available, you know, in hardcover everywhere, but uh there's gonna be a free e-book version starting in September, one year after the original publication date. And so I'm hoping that more people are gonna get the benefit of some of the stories of these great nonprofits that have learned how to use this and also the stories of where people have failed, because that's the way I think we all learn is by hearing about stories of our peers.

Carolyn Woodard

Yeah, no, that's great. And I will include all of these links in our show notes as well. So if you're listening and wondering where I can get that book without waiting for uh September, then we'll have that information to you. I just ordered your book. Right. Should have waited, but no, I um, you know, I remember you talking about it, but then I was preparing for this today and I was like, I need to read that book.

Jim Fruchterman

Well, thank you. Well, I mean, there aren't that many books on nonprofit tech in general. And my goal was to come up with a book on that was about how to do sort of the team building and the product and the revenue generation and the fundraising, the you know, the the the package of how to do it as opposed to um something that's aimed at most nonprofits who are just trying to figure out how to use it, right? Which is I think a bigger market. But but I think there's so many. I mean, you know,

Jim Fruchterman

When I started the first 10 years, I didn't know any other tech for good organization. And now there are thousands of organizations that have tech teams. And because if you're ambitious about social change, you're probably using technology to, you know, help a million people. Software and data is probably going to be involved in there somewhere.

Carolyn Woodard

Well, Jim, I just want to thank you so much for sharing your time with us today. It was really just a delightful conversation, and I feel like I learned a lot.

Jim Fruchterman

Well, thanks for having me on board, and I hope people go out and do great things with tech for good.