Episode 24: AI for Fundraising with Gravyty's CEO Adam Martel
In this episode of AI at Work, Rob May interviewed Adam Martel, CEO and Co-Founder at Gravyty. Gravyty is a Boston-based artificial intelligence company developing self-writing emails and other amazing artificial intelligence products to revolutionize frontline fundraising at nonprofit organizations. Tune in to learn how he started Gravyty, their breakthroughs on behavioral change, his best advice for implementing AI at your organization, and much more.
Rob May, CEO and Co-Founder,
Adam Martel, CEO and Co-Founder, Gravyty
Rob May: Hello, everybody, and welcome to the latest edition of AI at work. I'm Rob May, your host. I'm the co-founder and CEO at Talla. Our guest today is Adam Martel, who's the CEO and co-founder at Gravyty. They're a Boston-based AI company that develops self-writing e-mails and other AI products for fundraising at nonprofit organizations. Adam, welcome to the show and tell us a little bit about your background and how you started Gravyty.
Adam Martel: Thanks so much for having me, and I'm so excited to be here. I'm really happy to see you. Gravyty, we started that about two years ago, 2 and 1/2 years ago. My co-founder and I were getting our MBAs at Babson and I was working as a major gifts fundraiser there, raising money with the department for the school. What we found was that we were just severely inefficient in the work that we were doing. Now, we had these databases. Babson paid for all of this data, and it was siloed in the departments and the research department and some of the advancement services and IT departments. The front line fundraisers didn't have access to the data. More importantly, we really weren't using the data, taking advantage of it, to help us be better at our work.
I met this guy, Rich, in the MBA program. He was getting his MBA and he focused on, in his previous job, building machine learning algorithms that could predict which stocks were going to pop the next day. He was applying machine learning, trying to take all the data that they had publicly and making predictions about which stocks were going to pop the next day. I was telling him about the challenges I was facing as a front line fundraiser. What he said was, we could probably take a lot of the data that we had in our donors and apply some of the machine learning algorithms to be able to predict which donors were most important to focus our attention on, the same way he was doing that with stocks.
It was just an amazing process. We learned so much because we sort of found that we were spending a whole bunch of our time running primitive predictive analytics in our minds as frontline fundraisers, but most of the fundraisers I worked with didn't have any skills in analytics or quantitative skills. They were really having a hard time figuring out which donors to be in touch with. These algorithms gave us the ability to focus on the right donors. We built out these dashboards and we raised some money for it. We sold the dashboards and we found that nobody logged into the dashboards. That's sort of the perennial problem of, you have a tool and nobody uses it.
What we found was that frontline fundraisers, they didn't want a fourth screen. So, we unearthed this fourth screen paradigm where everybody has three screens in their life. They have their email, their calendar, and their CRM. They don't want a fourth screen to tell them what they do in the CRM, which tells them what to do in their email, which tells them what to do in their calendar. We pushed everything out into email and we quickly learned that not only could we predict which donor's most important, but we could actually self write the next draft of the email to that donor.
We built these great algorithms that had the ability to track the changes that the frontline fundraisers made to the emails. That gave us the ability to refine the writing style, which means that each individual front line fundraiser was now training their own writing style and we could get better with time and make better predictions about what they should be saying to the right donors at the right time.
Ultimately, we've proven that we can expand the workforce with artificial intelligence. I'm sure we'll get into that. It's just been a wild and amazing ride over the past couple of years.
RM: Your fourth screen thing is interesting because we've seen a similar thing here at Talla, which is when we're trying to work with groups that are doing customer support automation and stuff like that, some of them work in Slack or Microsoft Teams. Some of them work in a CRM or their ticketing system. But, they don't want another place to go. There's some things you can only do sometimes in your own interface. What we had to do was build a Chrome plugin so it can work in the browser. It can work in whatever system they're in.
I think this is the common model with AI tools. You start by fitting into an existing workflow, providing some small AI value. Over time, when people see the value, they're more willing to use your tool more and more in your own interfaces to maybe configure the tool and make it more powerful and all that. It's interesting that you had a similar experience.
You recently wrote an article for the Huffington Post about how nonprofits can use AI to improve fundraising. What are the things, you work with a lot of customers now. You guys are getting some pretty good traction. What are the things that you've learned, how big of an impact can you have on an organization? What are the key takeaways from your best, most successful customers?
AM: That article was fantastic. It was actually written by the Chief Innovation Officer of Salesforce. He was interested because Salesforce is applying artificial intelligence in sort of all that they do. I think every organization and every publication that's covered us, what they're most interested in is the impact on our customers and how they're using the technology. You just talked about the Chrome plug-in. One of the biggest insights was that we really, we can't get our users to install Chrome plugins because they're on the road so much.
There really is no Chrome plugin when you're on the road. What we found was that we had to use marketing techniques and the marketing email techniques to make the emails interactive. Everything happens real time. There's nothing to install. We found that barrier to entry was great if you eliminated it. The biggest learnings we've come across so far is that artificial intelligence, for us, is less about the back end algorithms and more about the behavioral changes at the organization. We're talking about applying AI to folks that, you know, they're in fundraising.
They didn't really go to school to be fundraisers. They're really relationship builders. They don't have a lot of the quantitative skills that a lot of the sales folks have and the sales ops have, but are still required to hit metrics, and it's a numbers game. If you can make a front line fundraiser more efficient, if you can empower them to play the numbers game, get some more donors and inspire more donors to make gifts, more donors will make gifts. The actual silver bullet is to find ways to get the front line fundraisers to use the product in a way that helps them directionally get out to the donors.
Behavioral changes was the biggest breakthrough that we've had in learning that even if the efficacy of our algorithms are right, even if we're 90% right in our algorithms and in our predictions about which donor's most important, it doesn't really matter unless we get the front line fundraiser to take action on those predictions. That's what the self writing of the emails was so important, saying, hey, we're going to start you on second base. What we learned was that really being able to apply the AI to the behaviors of these frontline fundraisers and augment their behaviors was most important.
At College of Charleston, what we found was that the expansion of workforce is the outcome of that augmentation. They have 15 frontline fundraisers and they're on pace to do the work of 39 frontline fundraisers. The University of Delaware, they have 20 frontline fundraisers. They're on pace to do the work of 90 frontline fundraisers. They just enter the $50 million donor that looks like it's going to close a gift. When you start talking about those numbers, the SAAS-based metrics, they just don't apply. Do you sell this thing per month, per seat, when there's a $50 million gift at play?
It just doesn't make any sense. You're really looking at human capital and you're connecting software-- the application of software, pulling it from a human capital budget because you're actually saying, instead of hiring 40 or 50 more fundraisers, you can make your current fundraisers and empower them to do the work of those folks. It's an amazing outcome that I think we're one of the first companies to really, really have nailed.
RM: I agree. It's something we've heard a lot of people talk about, but not a lot of people who've been on the podcast do successfully. We've seen a similar thing here at Talla, right? We had a customer who had 100 support reps and was going to hire 50 the next year and basically after a couple of months on the tool said, wow, we think you're going to save us 30 heads next year. It's probably a million dollars, million a half dollars for them.
It is interesting because the buyers are used to thinking about this from a software budget, when you plan out your budget for each department, you think about heads. You think about software. I think people forget, to your point, this should come from your human capital budget because you're either replacing heads or making the much more effective. It's a very different way to think about the buying process.
AM: I think it was Drew who said, on this podcast, he said they don't buy our software. They hire our artificial intelligence. And we're seeing the same thing where you're essentially hiring an extension of a person. And especially with frontline fundraising, like a lot of other tasks, these folks didn't go to school to be frontline fundraisers. The barrier to them getting up and running is between 6 and 10 months for them to actually be efficient in fundraising.
If you can take the people you have who are already trained and make them more efficient, it's an amazing extension of what can happen. Fundraising in general is going through a real crisis of talent. There's not enough talent, particularly when you look at these schools and these hospitals and these nonprofits that are in the middle of nowhere. There aren't that many professional, seasoned frontline fundraisers. So, what you do? That's where technology will truly accelerate the impact of these nonprofits.
RM: From the time somebody does a deal with you guys to the time they get up and running, what's the process like? Does it require a lot of training? Do you have to go in and spend a lot of time with people? Do you find people that don't adopt it quickly? If so, why is that? What's your process? What are your best lessons there?
AM: We have a 92% adoption rate, which is absolutely unheard of. I think that first of all, that's a tribute to us moving the entire system into email because you can't not do something with these emails. You're going to get them every single day, and they're going to change your behavior. We found early on that investing in customer success was absolutely something we had to do. Our first hire was a salesperson. If I had to start this company over again, our first hire would be a customer success person. I mean, all of our customer success folks are fundraisers.
We go in. We sell a deal. Usually it's a four to six month sales cycle on the deal itself. It's six weeks time to value. They're getting their money back in six weeks. They're getting their money back in six weeks, because there are gifts that are being closed that cover the price of the contract from donors that aren't being touched right now.
What we're finding is that customer success early on, making sure that the front fundraisers both know how to use it and understand what the use case is, has been a great learning. One of the most important learnings has been that use case. When we went out and we sold this first, we were looking for early adopters. I know there's this debate about whether we talk about whether we're AI or whether we're not AI, whether it's efficiency tool. Some people say talk about AI, some people say don't talk about AI.
We always talk about AI because we're trying to capture early adopters in the market who can lead towards the next cross in the chasm type of adopter. These early adopters, they have to have a rallying call. There has to be a use case, whether it's increased participation, whether it's increased revenue, whether it's increased activity from the frontline fundraisers, there has to be a use case at play when applying artificial intelligence. That's what our customer success team does best. We connect the front line fundraisers with both training and the importance of the use case. We use the entire management staff to push down the use case and then try to achieve a use case that's never been achieved before. That's what we've been able to successfully do with all of our customers.
RM: What have you learned developing a product in the AI space, dealing with data scientists, machine learning, and AI people in terms of your roadmap and how you think about it? Sometimes you don't know the outcome of some of your AI work and how well it's going to go. How have you changed over time sort of how you've thought about product management, maybe how you've structured your team? Have there been any major learnings there?
AM: When we first started the company, we went after the midsize organizations. Not the largest, because they have data scientists on staff and not the smallest because they really can't afford us. You go after mid-market then you'd be able to sell a lot of this because they actually need it. They don't have enough frontline fundraisers, have too many donors, have some money, but not an abundance of money.
What we found was that the cost to close a deal for mid-market was the exact same as the large markets, but the high markets obviously gave a larger ACV. We quickly learned that we had to be both an R&D company and a true AI product company. Developing AI products with an R&D spin was sort of how we crossed the chasm when it came to actually communicating that this was something that was really important. Then the customers became successful with the application of customer success. The biggest learning there was that all AI companies are essentially R&D companies. If you confuse yourself with the SAAS metrics, you're going to lose that R&D focus. And now you have to keep developing and you have to keep learning.
RM: It's very interesting. It's one of the mistakes I still see a lot. Investors are starting to come around in this space. I've seen a lot of your smart venture capitalists that look at AI. They've been doing bigger deals for funding rounds. They've been reserving more capital. I'm very bullish on the fact that maybe in another two years, the vast majority of VCs and angel investors will really understand how these companies are built. It's part of what kept people out of hardware for a while, too.
A lot of our listeners are executives at companies that maybe aren't super technical in terms of what their final product is, although all companies are obviously relatively strong tech companies now. They hear a lot about AI. They read a lot about AI. They're listening to this podcast to learn. What's your advice to somebody who came up, they studied business management, marketing, accounting, whatever. They're a senior executive at a company now. They use tech. They understand some simple concepts. But, they're very confused by AI. Where should they go to learn about this? What should they pay attention to? How do they separate the crap from the good stuff?
AM: That's such a good question. There's so much fakeness out there and real marketing spin. Of the best pieces of advice is to worry less about the term of artificial intelligence and worry more about the actual outcomes that are being promised and make the decision based on whether you believe that the AI company can deliver those outcomes. For some reason, we keep confusing AI as the outcome. AI is the “how” to get to the outcome. For us, can we get the frontline fundraisers to be more efficient to contact more donors?
Whether it's AI or automation or whether we have 500 people in the back that we can build a successful company on that are actual personal assistants, who cares? What truly matters is that we're scaling the ability of these frontline fundraisers to get to more donors. My suggestion would be true for all of these folks, really evaluate the products. You know, take demos. Take the time to look at this stuff because most of it's going to be much different than each other. You're not going to see many trends. Try to evaluate whether the outcomes are worth the investment, both with time and money. Don't confuse AI as the outcome. Just because you have an AI strategy doesn't mean it's good for you.
RM: That's great advice. When you look forward at the future of AI and where it might go over the next five to seven years, what are you most excited about? It can be from a technical perspective, from a market perspective, or anything?
AM: We just released an AI and advancement council. It's 16 of the top leaders in fundraising. We took some of our early donors, we took some folks we really respected in the fundraising space, folks that had the ability to actually make change and the folks that have assistance, folks that have a budget, put them all on an AI and advancement council. We just had our first call a couple of days ago. What we discussed was what is the future of the space and what's the point? Everybody got on. They introduced themselves.
The common thread there was that, this is the first time that people that have been in this field for 20, 30 years have seen something new. It really wasn't since the web that they had seen something that could be as transformational as AI. That's what they all sort of said in accumulation was that we want to be at the forefront of whatever this wave is. Everyone's calling it AI, but whatever it is, we want to be at the forefront of it because we know that the way we're doing things doesn't correlate to the outcomes that are expected of us from our boards and our leaders.
Specifically, I don't know what the next five years, 10 years look like. I know that every person will probably have some type of AI assistant or assistance in their life. We already do. Most of us already have some type of AI assistance. When it comes to enterprise and work, that will be redefined by technology and our ability to be great at our jobs. The definition of greatness will systematically change. That's exciting.
RM: I think the thing that you mentioned about how you have a group of people that haven't seen a new innovation like this, there's so many industries that are going to go through that, where they're like, wow, this is really new innovation for my industry, the first thing I've seen in a long time.
AM: Every startup should have an AI in whatever council. I mean, there's no reason not to put a council together and get some of the smartest folks and the industry leaders.
RM: What do you wish somebody would build in AI to do that hasn't been done yet? As somebody who uses some of these products, obviously you know a lot of people in the AI space because we kind of all have to hang out and commiserate together and everything else. Is there a gap? If somebody is sitting out there and they're thinking, like, “Oh, man, I should start an AI company” is there a gap you see of, “If I wasn't doing Gravyty, I would go do this?”
AM: The question isn't what could you do. The question is, I mean, what couldn't you do, right? It's like, we're at the tip of this iceberg. I remember you and I spoke a year and a half ago, two years ago about how great this will be. And even the progression and the progress that's been made in the past year and a half, I mean, there's huge opportunities in CRM itself.
You know, we're in the fundraising space, but looking at sales, I mean, every small business has a sales team. Whether it's chatbots, conversational AI, whether it's something else completely, there are opportunities there to revolutionize not the process but the outcomes of what's happening. I think that's what's most exciting. Before we start another company, we'd be looking at another way to apply AI to really influence outcomes.
RM: I had a discussion the other day with somebody about taking a bunch of AI and tying it together with a bunch of marketing and sales and pipeline systems. If somebody had a relatively well-defined and mostly online buyer's journey and you could tie-in from Google AdWords to Stripe, for example, could you almost build a push a button get a customer kind of thing, where it's like, hey, you quote them a price based on your prediction of what it would cost to get them a customer a certain amount, and then you take the risk with your system on the back end of getting the customer cheaper than that, right? Somebody will build that in the next couple of years.
AM: Absolutely. What's interesting about the nonprofit space is that we have all that data. I mean, these are donors who went to the organization. They had some type of affiliation with the organization, whether it's a higher ed or hospital or nonprofit. They've been there. They've been to events. We know a lot about them. They want to be engaged. The question is, can you engage them personally and in really smart ways? You plug that into sales, there's no question that's going to be revolutionized.
RM: Very interesting. So, looking ahead, 2019, you have any predictions for AI for the coming year that you want to share?
AM: Yeah, a couple. I think specifically for us, we're going to be focused on outcomes. We had a huge year announcing partnerships, announcing new products. Some of the partnerships we announced, like the-- we announced an iWave partnership, which gives us-- in the fundraising space, there are companies that have all of this data on donors that predict wealth. And that's being plugged right into the artificial intelligence. There are some really great opportunities there, I think, for data companies, for CRMs, and for AI companies to all partner together to provide a verticalized outcome.
I think 2019 will be defined by verticalized artificial intelligence, AI that specifically does something for a very specific person at a very specific organization, and enables them to be better than they are. So whether it's a frontline fundraiser or whether it's a dean, whether it's an educator, whether it's a machinist, whatever it is, AI will redefine that space. And verticalized AI will accelerate the pace at which we're moving. I think that will define 2019. It will be the bringing down of the silos of the different organizations.
What do you think about this article that recently came out in Harvard Business Review about AI adoption and the fast follower strategy? You had Tom Davenport was one of the articles and-- was one of the authors. I forget who the other guy was. But part of their argument was fast following in technology has been a valid strategy for the last 30, 40 years. Let somebody else work out the kinks, work out the bugs. What do you think about that with respect to AI, right?
If I'm looking at Gravyty and I say, I'm going to wait three years and then when all the kinks are worked out, I'm going to do it, the people that have been using Gravyty for three or four years, are they going to be so far ahead? And will they have changed the way that they work and the data they have? Will it be hard for the late adopters to catch up? Or do you think fast following is still a legitimate strategy?
Tom Davenport is one of the grandfathers of the space. He's actually a professor at Babson. My co-founder teaches with him. Richard I had this conversation a couple weeks ago. I mean, the fast follower mentality just doesn't work in AI because AI is about acceleration. It's not about redefinition of what it is. It's not about accumulating data for data's sake. It's not about the siloing of data. It's really about the acceleration of the outcomes.
If you consider two organizations, two nonprofit organizations, one that implements AI and accelerates, you take a U Delaware that becomes 90 frontline fundraisers and you compare them to an organization that only has 30 frontline fundraisers, the one with 90 is going to win and they're going to capture enough of their donor base where if you're competing for donors, for those same donors, there's an unfair advantage for the company with AI. I don't think organizations will catch up if they don't adopt AI early.
I think that there's going to be an arms race sooner rather than later for who can adopt AI. I tell the story sometimes about my team was asked to pitch at a really well-known university out in the Midwest. We went out there. There were four of us that went out. We pitched the entire department. There were 20 people in this pitch. We had a great pitch. The VP called me a couple of days later and he said, hey, you know, we love the pitch, but unfortunately, we're going to wait because you're a little bit early. We're still not quite sure where AI is going.
The next day, I got a call. It was a woman on the phone. She said, is this Adam Martel? I thought, uh oh. I'm in trouble. I said, yes ma'am. It is. She said, well, I'm the president of that organization you just pitched. This is a really big organization. She said, let me tell you quick story. She said, “I was at Harvard in the '90s when they were trying to decide whether they should develop and implement a website for giving. I advocated for it and I was on the right side of history there. And we're going with you. And we're going to be on the right side of history here.” If you look back at history, the folks that are early adopters are always the winners. I mean, there's no situation where early adopters don't win. Fast following just doesn't make sense.
RM: Very interesting. Last question, and we ask this to a lot of people who work in AI. There's been, although it's slowed down a little bit, I'm talking to our producer here, Alyssa. The press around the Elon Musk Mark Zuckerberg debate has slowed in the last six or nine months. You've heard less about it. Maybe we won't ask it as much in 2019. We're still doing it for now.
You know, they had this big, public spat about whether we should be worried about generalized AI taking over the world, whether AI safety is something we should be incredibly concerned about, or whether we have time to think about it. You know, how worried are you as somebody who works in that space? Where are you on the Musk versus Zuck spectrum? What are your thoughts on the potential dangers out there?
AM: AI is defined by the use case. Bad people with AI will create bad AI. Good people with AI will create good AI. We are a team of good AI folks. And we're trying to create good AI, as are you. You just don't know. I mean, you just don't know who is going to do what with what. But I think the responsibility of technologists is to develop artificial intelligence in the most ethically possible way to do the most good for humanity. I think that if you compare the best good with the best bad, the best good will always win. That's what I'm putting my money on. I'm going with Zuck.
RM: Awesome. Well, Adam Martel, thanks for coming on. If people want to find out more about Gravyty, or if they're interested in seeing a demo or becoming a customer, what's the best way to find you guys?
AM: Visit us at gravyty.com, or you can email me directly at firstname.lastname@example.org.
RM: All right. Thanks everybody for listening. If you have guests you'd like us to see on the show, questions you'd like us to ask, topics you'd like us to cover, email email@example.com and we'll see you all next week.