Why AI and Machine Learning Fail to Help Fundraisers Build ‘QUALIFIED’ Major Donor Portfolios (or Caseloads)

AI and machine learning are the new buzzwords in the prospect identification field proposing yet another solution to the challenge of building caseloads of major donor prospects. But can they really help you fill a caseload with qualified major donor prospects better than other methods?

Let’s start with understanding what we’re talking about.

What Is Machine Learning and AI?

Machine learning attempts to use computer algorithms to study all available data and use it to make predictions and models about future behavior. But in addition to that, it builds on its previous efforts by incorporating new data as it comes in. It learns, based on ongoing human behavior and data.

You have most likely experienced machine learning without even realizing it in services like YouTube, TikTok or Netflix and many social media platforms that offer a steady stream of things to click on based upon your previous behaviors and interests.

The algorithm ‘learns’ about you based on your ongoing actions within the platform, as well as on previous data it has already collected about you from elsewhere.

How Does AI with Machine Learning Get Used to Build Caseloads of Major Donors?

Generally speaking, wealth screening vendors attempt to analyze two sets of data – wealth indicators and philanthropic indicators. They do this initially by analyzing your data (mostly your donors’ recency of giving, frequency of giving, and the dollar amounts of those gifts).

Then they feed more data into the mix including publicly available wealth indicators such as political contributions, stock ownership, real estate, and business affiliations. All of these are signs of wealth and a capacity to give.

Next they’ll include philanthropic indicators which include volunteer experience, the amounts of their prior donations to organizations like yours as well as to others, board involvement, and foundation trustee positions.

Their AI uses all this data to determine two metrics: Affinity to give and wealth capacity. Then, it combines these to produce a new ‘predictive’ metric: Likelihood to give.

What’s the Problem with AI and Machine Learning for Major Gift Caseloads?

While AI and machine learning certainly have some advantages over traditional wealth screeners and RFM data, they still suffer some of the same shortcomings.

Insufficient Data

All the models and predictions generated by machine learning are based on past data and previous behaviors. Its output can only be as good as the data that it has to work with. And most, if not all, of that data is quantitative, not qualitative. Qualitative data is simply better.

However, if you are not engaging your supporters with a donor-driven system that collects qualitative data, AI cannot know why a donor cares about your mission. It cannot know the values that drive a person to give or participate as a volunteer. It cannot understand the role another person played in the journey of a supporter to become a donor to your nonprofit. These are the foundations of donor motivation. They drive giving at a high level much more than affinity and capacity (both based on quantitative data).

Plus, when it comes to major donors, usually the sample sizes of quantitative data are too small to draw valid statistical conclusions. Once you have a set of major donor prospects, how do you qualify and cultivate them? The AI supplied by wealth screening vendors will not be much help here, because each donor is different.

But is there a way to qualify donors without talking to them? No, AI can’t do it. But as you’ll see in a bit, there is another tool that can.

Outdated Data

Past information (which is mostly gathered from publicly available sources) doesn’t always align with present reality. In most cases it becomes outdated soon after it has been collected.

A person may have recently lost their business or gotten divorced, and their net worth has cratered. Will the AI know about that? Or will you traipse into a meeting all excited about this prospect’s wealth capacity and likelihood to give, and then get blindsided with the bad news?

On the flip side, other supporters may have much more wealth than the data indicates. Perhaps they just eliminated some debt, sold a business, received an inheritance, or landed a big high-paying job. AI’s data sources won’t know that, which means AI won’t know it either. But that person has the wealth capacity to make a major gift.

Inaccurate Data

A very academic and heavily mathematical study came out in 2022 studying various methods of using AI and machine learning to identify major giving prospects. I wouldn’t recommend reading most of it – unless you have a math degree.

Its conclusions actually revealed the shortcomings and flaws of AI and machine learning, even though the study author seemed to consider them a success. But that’s because the study author doesn’t work in fundraising. This is a tech and math wizard, and their software “works” in the lab.

For example, depending on the mathematical model they used, AI software was able to accurately identify between 60-85% of major donors and non-major donors.

Again, the study’s author considered this a success.

But is it?

Going with their best numbers, that means you are mis-identifying 15% of your major donors. So, you would be pursuing prospects who cannot and will not give major gifts, losing about one sixth of your time and resources on them.

That’s not good.

And, the software also failed to identify a comparable percentage of actual major donors, putting them in the non-major donor category.

That’s lost money. These are people who could and would make major gifts if properly cultivated. But AI didn’t even identify them as prospects, so you’re not reaching out to them at all if you rely on the software to build your caseload.

What AI Does Well

AI cannot know the qualitative data that is essential to conduct the best possible pre-qualification (for outreach), cultivation and solicitation of major donors. This includes things like values, personal stories, history with your organization, desires, concerns about giving, preferences for different giving methods the actual donor may not even know about, and so many other things that can only be learned through personal conversation.

However, AI can provide a good starting point. If you are just starting out with major donor cultivation and want to gather an initial caseload from your database, AI will speed up that process.

Plus, it can keep track of the donors you are reaching out to, and improve its processes and recommendations about how to work with them over time.

AI also can do well at large database tasks like identifying potential monthly donors, or finding segments that might be interested in giving to particular fundraising campaigns.

But one thing AI cannot do is qualify individual donors. That requires engagement.

What About ChatGPT?

Now, you can talk to AI like a person. In early 2023, this new service took the world by storm. But does ChatGPT help identify major donor prospects?

In a word – no. That’s not really its purpose.

I tested it by asking ChatGPT, “How can I identify major donors from my charity’s database?”

ChatGPT produced a 7-step plan:

  1. Select a great donor database
  2. Perform prospect research using publicly available data
  3. Clean up your donor database
  4. Look at past giving
  5. Consider event attendees who have capacity to give more
  6. Do the same with auction participants
  7. Look at your annual fund to find more major donor prospects

As you can see, this isn’t very helpful. It’s actually recommending a plan far inferior even to what AI software such as we’ve discussed in this article can do. And it’s very general advice. No details or recommendations for how to actually do it.

But let’s try a more specific question. I asked ChatGPT this follow-up question:

“How do I qualify major donors once I have identified prospects?”

ChatGPT recommended looking at wealth capacity, donor affinity for your organization or mission, and their interests and motivations. Then, it recommended tailoring your approach to each donor based on that.

So, as you can see, ChatGPT isn’t built for what we’re talking about. It can recommend action plans. But it is not designed to actually help you do the work of identifying or qualifying major donors.

A Better Method for Building Caseloads with Qualified Prospects

Again, qualitative data is much more powerful than quantitative data if you want to accurately identify and qualify major donor prospects with a near 100% degree of accuracy.

You need to know their current life reality in terms of finances, life situation, goals, desires, interests, values, and dreams. You need to know something about the story they are trying to tell about their own life, and how you can help them tell it better through transformational giving.

And you need to know all this before your gift officers reach out to them so no one’s time gets wasted.

AI cannot deliver this level of qualification, because the data on which it depends remains bound by the past. It can only make assumptions to about 80% accuracy, according to at least one study.

And hitting 80% is no good. You want to hit 100%. You want your gift officers to only make outreach to qualified donors.

MarketSmart’s software was invented to do what AI, machine learning, and other non-interactive methods of donor identification and qualification cannot do – find and engage potential donors, at scale, using automation, until they are ready to be reached out to by a gift officer.

Our software automates these earlier interactions so that actual people only need to get involved once the ‘digital body language’ from the prospect indicates they are ready.

For a qualified donor prospect, you’re looking for five things:

  1. The timing is right in their life to make a gift
  2. They have an emotional reason to give
  3. They have the wealth capacity to make a major gift
  4. There is a history of past engagement with your organization
  5. They have given their permission for you to reach out

When you have all five of these in place, a gift officer can reach out to a prospect with a high degree of confidence. And our software engages donors over time, keeping you apprised of their progress. When all five of these attributes are present, the software flags that person as being ‘outreach-ready.’

The best part is, our software can engage thousands of donors all at once.

This beats AI, because it produces zero false positives – donors you thought could give big but who can’t. It also misses far fewer false negatives – ones you didn’t think could make big gifts but who actually can.

How well does this work?

Well, we offer a 10:1 ROI Guarantee. So we’re pretty confident it works and we have the track record to prove it, or else we couldn’t offer such an extreme guarantee. You will generate at least ten times the cost of our software in new revenue. Click the button below and schedule a free demo to see how we do it.

Schedule My MarketSmart Demo


Related Resources:


Notify of

Inline Feedbacks
View all comments
Get smarter with the SmartIdeas blog

Subscribe to our blog today and get actionable fundraising ideas delivered straight to your inbox!