The Unexpected Wisdom of Serendipity

Have you ever been in a situation where it feels like the stars have aligned and you are exactly where you were meant to be? 

It’s happened a few times in my life. 

One time, when my husband and I were moving from Washington to Oklahoma, we stopped off in Yellowstone National Park to see Old Faithful erupt. After barely getting there in time, we decided to stay and watch it a second time. As we were waiting for a second eruption, a voice called my husband’s name out of the crowd. It turned out to be his entire extended family from Norway who just happened to be visiting Old Faithful at that exact same day and time! Oddly enough, they hadn’t ever met in person before; they only recognized him from his Facebook pictures (it helps that he is a 6’4” tall red head).

And the craziest part of that whole experience: we weren’t even supposed to be there at that time. We had left late and were running late. Had everything gone according to my plan, we would have completely missed them.

That was a moment for me where I knew I was exactly where we were supposed to be. 

So what does that have to do with boodleGPT? Well that is the exact same feeling we have here at boodleAI right now, as we take advantage of everything we’ve done previously to launch into the realm of Predictive and Generative AI. 

Everything we’ve done up to this point has been preparing us for this moment.

The Road (MUCH) Less Traveled

As we talked about it in our last blog post, there are three things a company needs to build a Middle Layer AI :

• A Unique Data Set
• Rare Talent to train models
• Power AI/ML and data infrastructure

You might as well be describing what we’ve spent the past few years building. As we’ve served the nonprofit and business communities, we’ve built out our proprietary database of 35 billion predictions for virtually the entire adult population of the United States. 

What insights are provided by boodleAI’s predictions? Two words: a lot.

Why boodleAI’s predictions are different

These predictions are unique in that they are focused on long-term, affinity-based signals rather than short-term behavioral signals that deprecate almost instantly. (In other words, we base our predictions on who people are rather than what they do.) While most of the data industry sped up, we purposely slowed down and focused on how to describe who people fundamentally are. We found (both in testing and through customer validation) that these long-term signals are much more predictive of extended future behavior than the short-term, intent-based signals available on most platforms.

Why customers trust boodleAI’s predictions

There are four reasons customers have confidence using boodleAI’s predictions for sales, marketing, and fundraising:

1. First, when we build our prediction models, we hold back a certain portion of the data used to train the models and use that “hold back data” to validate that the predictive model is truly predictive.
2. Next, we test our predictive models against millions of known data points and only once a predictive models shows that it can perform well compared to known results do we release the model for use by customers.
3. Third, our predictive models have been validated across the commercial and nonprofit industries with hundreds of clients who have used boodleAI’s predictions in donation campaigns, event coordination, sales forecasting, and targeting/personalization. We have countless stories where our predictive model not only met, but exceeded customer expectations. However, there is always room for improvement, so we continuously analyze every piece of feedback we receive to improve the accuracy and reliability of our models.
4. Lastly, if our predictions don’t meet expectations (which occasionally happens), we take full responsibility and go above and beyond to understand why and what we can do to improve results in the future.

What makes boodleAI’s predictions great for an LLM

Because each of these predictive models were generated using machine learning techniques, they have 100% coverage across all 250 million+ adult Americans. This gives us a very large, very specific dataset to layer on top of a foundation LLM like GPT-3.5 or GPT-4.

boodleAI’s predictions: Fuel for a better LLM

I’m going to generalize for a second and use ChatGPT as an example of all foundation layer LLMs. Part of the magic of ChatGPT is its vast knowledge of lots of things–that’s how it was trained after all. The training set for ChatGPT is huge and covers as broad a range of topics as possible. However, that’s also its main weakness: it’s vague. And that makes it difficult for it to give a user specific, actionable answers related to a particular industry or vertical.  

In contrast, our dataset is full of actionable, relevant, and specific predictions about individuals. (And we are in the process of creating actionable and specific predictions relevant to specific industries.) However, it’s always been a struggle to get boodleAI’s insights into the hands of the people to whom they are most useful because the data is not always user-friendly and requires contextualization to fully understand.

That’s why the combination of our data and a LLM like GPT-3.5 or GPT-4 is so thrilling. We are solving the fundamental problems each of us have. boodleAI gives an LLM like GPT-3.5 or GPT-4 the ability to provide user-specific, actionable insights particular to an industry or even an organization. An LLM like GPT-3.5 or GPT-4 enables a user-friendly interface to access and apply boodleAI’s insights while also providing contextualization that helps users fully understand and use the insights.

And that’s not all!

In our next blog post, we will explore how the combination of Predictive AI and Generative AI can unlock a whole new generation of Middle Layer AI. Spoiler alert: boodleAI already has the infrastructure to do it!