Cheap (or free?) AI for Nonprofits?
Updated: Oct 24
At Dreamforce last month I was excited to once again hear people from Salesforce.org (SFDO or “.org”) mention Einstein, Salesforce’s name for their machine learning offerings, and how they might be used by nonprofits.
You see, several years ago, when Einstein was first announced, I tried to pin down Salesforce.org executives to find out if/when Einstein would be available to nonprofits, how we might use it, and what kind of discount we could expect when pricing was announced. I got exactly nowhere back then. List pricing for Einstein wasn’t released yet and nobody at .org would commit to what kind of discount they would offer nonprofits in percentage terms. (Not my first or last tilt at transparent pricing.)
In the intervening couple of years Einstein has been front and center of messaging but I will say I have barely heard a word outside of Salesforce’s own marketing about people actually using it. And to add the confusion “Einstein” is now applied to several different products, such as Einstein Bots, Einstein Activity Capture, and more. You’ll sometimes hear people (including me) joking that Salesforce is just going to prepend “Einstein” to the name of any product they want to imply is “smart,” whatever that might mean.
The Einstein product that I’ve been keeping an eye on from the beginning is currently called Einstein Prediction Builder. This is the one that most closely matches what I think of when I consider how machine learning or “AI” could be applied for nonprofits. It’s the tool that should allow me to one day, for example, to look at potential matches of mentors with mentees and find those that are more likely to produce good outcomes. Or perhaps I could point Einstein Prediction Builder at a set of students and their program engagement data and determine who among them is in danger of dropping out of the program?
So imagine how interested I was in the Nonprofit roadmap session to hear that Salesforce.org has actually put some effort into making it easier for nonprofits to use Einstein Prediction Builder. That’s potentially really cool!
If we can figure out what it means.
And I’m not trying to knock SFDO here–I actually think their materials were about as clear and transparent as a high-profile product announcement was ever likely to be. But Einstein is complicated, Salesforce licensing is complicated, and what Salesforce.org built is a little challenging to sum up in just a couple of sentences. So let’s start with some gratitude: Kudos to SFDO for trying to get Einstein Prediction Builder into the hands of nonprofits! There’s potential there for some interesting wins.
The Announcement: Einstein for Nonprofits
(I don’t actually think that Einstein for Nonprofits was a new announcement at Dreamforce, but it was a nice high-profile reminder that this is something SFDO is counting among its big releases for 2022.)
Let’s break down the dense language of the linked help page to understand what this is. It’s really two things:
1. It’s Einstein Prediction Builder (EPB)
That’s not really a special product for nonprofits. Any org can turn on EPB in “Try Einstein” mode and build a couple of predictions and even activate one of them for free. So this is a bit of repackaging of Salesforce platform capabilities to let nonprofits understand that we get them as well.
2. It’s also some additional sweetening by Salesforce.org
a. First of all, Salesforce.org has built three prediction models for us. That’s actually pretty cool, as building prediction models is the difficult and confusing part of using EPB. In this case, SFDO has created predictions for whether a contact in your system
is likely to become a first time donor,
is likely to become a top donor, or
is likely to become a recurring donor.
These predictions are all based on the NPSP donation rollup fields. So as long as you’re using NPSP and its customizable rollups in more-or-less the way they were designed, these predictions should work for you out of the box. Neat!
Another benefit of these predictions is that they serve as examples of a 100% nonprofit-specific use case and exactly how you would implement it. I’ve done the Einstein modules on Trailhead and come away with a vague sense of what I might try to build for a nonprofit but there was a good deal of fuzziness there. Seeing an actual prediction and then actually using it against real data is the kind of learnin’ I need.
b. SFDO has also created “backup models” for those orgs that don’t have enough data. One of the limiting factors of Einstein (or any machine learning) is that you have to feed it a whole lot of example data so that it can analyze and look for patterns. An Einstein prediction won’t even run if you have fewer than 400 example records and can point to a good chunk of both positive and negative outcomes. If you’re an organization that’s new to Salesforce or doesn’t have hundreds of donations imported from an old system you wouldn’t be able to build an EPB prediction at all. But SFDO has worked to gather a big dataset and build the models you’ll need so that you don’t have to start from scratch and you can use EPB before you have enough of your own data to train it on.
c. An Einstein for Nonprofits app that can make it easier for you to find the EPB settings, the backup models that SFDO provided, and links to resources.
d. There are also some Lightning Web Component cards that you can put on a contact page to display the results of the Einstein predictions. [Can I be honest and say that I don’t think these cards are really worth much? Sorry–I’m callin’ it like it is. We could replicate this display in five minutes with the Related Record Hack and we’d have better control of what we are displaying.]
You’re probably wondering what Einstein for Nonprofits costs. [At least, you’re probably wondering this if you’re cynical, like me, and assume that we can’t be getting anything worthwhile for free in this world.]
I’ve got good news on that front: Einstein for Nonprofits is free. Everything I listed in the section above is totally free to nonprofits. One active predication is actually free for any org in the first place on the Try Einstein level of EPB (more on this below), so there’s no free platform feature being handed out here. But SFDO is throwing in those pre-built predictions and the backup models to train them, as well as those pretty-but-inconsequential contact cards. That’s actually a pretty nice deal, I think.
And if you want to upgrade from the Try Einstein level that only allows one active prediction, you can get an Einstein Predictions license for $225/year, a 75% discount from the regular price. An Einstein Predictions license is a feature add-on to one of your existing user login licenses. You need that feature license only for the person that is going to create and manage your Einstein Predictions. Everyone else will be able to see the results. With an Einstein Predictions license you can also activate up to 10 predictions (and build up to 20.)
Should I use this thing?
If you’ve been paying attention, I bet you have one more important question about Einstein Prediction Builder and Einstein for Nonprofits: Should I use it? (And, if yes, how?)
Well I’m still working on investigating that with the help of some of my clients that have a good amount of data in their orgs. I’m going to make that the subject of another post as soon as I can. Tentative title: My First Einstein Prediction.