An animal shelter in southwest Florida boosted its giving over the summer after it used artificial intelligence to quickly craft tailored messaging for a campaign.
The Suncoast Humane Society typically doesn’t run summer fundraising campaigns, but the organization wanted to see if A.I.’s ability to generate content could help it quickly create an involved project. The campaign followed a dog named Max and a cat named Whiskers through their journey from street to home with the help of the society. The campaign featured direct-mail appeals, emails, social-media posts, and a weekly blog from the point of view of each animal. The direct-mail and email pieces, which linked to the blog posts, were tailored to the two groups it was trying to reach — lapsed donors and people who had adopted pets — also using A.I.
“We were trying to re-engage donors that had lapsed,” says Rafael E. Robles, Suncoast’s chief development officer. “That was the driving force.”
Suncoast enlisted Josh Hirsch, a senior strategist at Soukup Strategic Solutions to help with the campaign. Hirsch says the A.I. helped build personas of lapsed donors and supporters linked to the society’s VIP — very important pets — program that encourages adopters to become donors.
The organization customized the GPT product, a service of OpenAI, the company that created ChatGPT, by training it on “organizational documents, past impact reports, past blog posts, past schema, so it understood the brand voice and tone,” Hirsch says. The customized GPT was accessible only to the humane society and didn’t share organizational data back to the larger GPT, Hirsch says.
A series of emails introduced Max and Whiskers and shared their back stories. “Whiskers was skittish and afraid of humans at first. And same thing with Max,” Hirsch says. “We were able to show how through the love, care, and the resources available through Suncoast Humane Society, they were able to find their ‘furrever home,’ and now they wanted their other friends to have that same opportunity.”
Hirsch, who finalized the blog posts, says the GPT output was a pretty good first draft, because of the document training, but he still had to revise and add the human touch. “From an efficiency standpoint, I was able to generate and do a lot more in less time and get this campaign moving a lot faster,” Hirsch says.
Robles wasn’t sure how much the drive would raise because the group was reaching out to lapsed donors and adopters, whose giving was inconsistent. He hoped for $25,000 as a stretch goal. By September, Suncoast had received 197 gifts from 194 donors, totaling almost $37,000. The average gift was $186.
“We were seeing engagement from donors who hadn’t given in three, four, sometimes five, six years,” Hirsch says. “And they were giving two or three times their past donation.”
Lessons Learned for Next Campaigns
The campaign was a learning curve for several reasons. One, the humane society has incomplete records. The organization has email and mailing addresses for some supporters, but for others it has only one or the other. So the society knew some direct-mail recipients also got emails, but if a donor was on the email list and also loaded as a separate entry for direct mail and got both pieces, the society wouldn’t know it. The society is seeking more complete information from donors so the database can be more accurate.
The organization is also reconfiguring its database so fundraisers can add more fields that will help them better personalize donor appeals.
“We’re trying to input as much of those data points as we can. For example, folks will write little messages on the return card, like ‘I adopted Jamie 12 years ago. She was the love of my life. And I miss her terribly. She’s no longer with me,’” Robles says. “Now we know the name, we know how long ago, we know the emotional connection to us through Jamie. And that then helps craft the story and the ask moving forward.”
Hirsch says it’s important for nonprofits that want to craft more personalized asks to have defined fields to put this kind of information into their data-management systems. He says when people don’t have those fields, they put the information in the notes section.
“Notes fields don’t help, in my opinion,” Hirsch says. “When you’re really trying to pull reports, that’s a little more difficult as opposed to having a set variable field where that data is being tracked.”
At first, using A.I. to create content felt a little out of Robles’s comfort zone. “I’m a traditional fundraiser. Direct-mail pieces have been in existence forever and ever, and we swear by it,” he says. “This was a collaborative between the old and the new.”
But now that he has the results, he’s ready to see what else A.I. can do. “I thought it was going to be all or none,” Robles says. “It turned out to be a synergistic collaborative between what we used to do and what we can do now with the A.I.-generated content.”
