Be wary of the AI magic bullets

I came across this announcement in my LinkedIn feed. I had to read it three or four times to make sure I was getting this right.

It’s becoming increasingly common that a lot of our fundraising technology partners are not only working to include some element of AI into their products, but they also have to declare why it’s better than competitors’ products. The danger is marketing frequently transcends the boundaries of reality (gasp!).

Case in point, BB is in no way shape or form an AI-native company. Nor is Raiser’s Edge an AI-native app. There is a little tug of war going on between the fundraising technology sector on who can get helpful products to the market first by either making them simple and cheap which means AI bolted on, or they are being slow and deliberate and being AI native. The former has to come with a little bit of nuance though. Some technology offerings are claiming they are “AI-powered” (what does this even mean?) because they have enabled a window that accesses ChatGPT. Nothing impressive about this. Some companies are able to build AI features within their own platform, but without completely rewriting their application to be truly AI-native. This nuance is important. Blackbaud’s offerings Fall into this category. They have not completely rewritten their applications to be AI-native from the ground up. But they were able to go a bit deeper than simply tacking something on the side. Buyers need to know this especially if they are under pressure from leadership to bring AI into their business practices in the most meaningful and consequential ways possible.

The quality of AI will be measured in part by how deeply integrated it is into the applications’ services, or even better if the AI is the service. Let’s put this another way. If you were to turn off the AI features in Raiser’s Edge, what core functionality would break? You see? It’s not an AI-native application. The basic functions of the application still work just fine.

From the perspective of my personal passion, data analytics, this makes a huge difference. When AI is bolted on after the fact, the results will largely be deterministic. The user does not have much ability to build the data model, interrogate the results, and refine the output. In the best of worlds, AI should be probabilistic, which doesn’t lock the user into a single experience with a single set of outputs.