Leveraging AI for Donor Analysis
Yesterday, I had the chance to present a breakout session at the Minnesota Council of Nonprofits Fundraising Conference in Brooklyn Park, Minnesota. What a blast! It’s always inspiring to hear about the great work our nonprofit organizations undertake. I am energized by the fantastic development professionals I got to meet yesterday. I am grateful for their work.
I’m happy to share my presentation here. I don’t like to read off of slides so the full context may not come through by reviewing the slides alone, but you can certainly get the idea of what I covered. I also shared a (not very pretty, I really have to partner with a graphic designer one of these days) handout you can take a peek at as well. Here’s the rundown:
We began with my favorite framework for thinking about AI, which is that it’s similar to an eager Intern. It’s happy to help, but makes mistakes; knows some stuff but can’t replace our own wisdom. I then walked through some donor segmentation strategies that are common in fundraising and asked participants to think about their own shops and how they could take the next step in segmenting their donors to improve the effectiveness of appeals and campagins.
With that in mind, we then spent a little time talking about data privacy. I shared some common techniques for anonymizing, coding or generalizing information and emphasized the importance of never uploading any personally identifiable information with our free, online tools. I shared an example of how one could translate a series of real information (Luis@aol.com gave us $2,500 last year, is married, and is vice president at the local bank) into a series of codes (890257, 2.5, 1, A) which would be meaningless should anyone every be able to get a hold this data. This served as a reminder for the importance of developing an AI and data privacy policy even in a small shop. I demonstrated how AI tools can help us think about critical questions to ask and even give us the first draft of a policy to begin this work.
Next we were able to actually dig in and use the tools! I had a static example where I used some qualitative data (in this case, a series of donor call notes) and asked ChatGpT to code these as high, medium, or low prospects. The system did a pretty good job with this task. Then I did my high wire act and uploaded some (fake) quantitative donor data and walked through a series of analyses. We explored how the system might generate a recommended gift amount or help us identify unexpected trends.
We ended our very brief time together by having everyone in the room explore their favorite AI tool and, using their own context, ask the AI systems to help them. One participant said her organization wanted to explore how to reach their volunteers differently and started a conversation about how they might test different campaigns for volunteers to see if the results were beneficial.
There’s never enough time to dig into all the nuances in these sessions, but it was clear participants left with some new ideas for how they might leverage the AI tools at their fingertips to explore new strategies. It was a fantastic end to a rich day of learning!
