AI Strategy at Rhyze
May 25, 2024

What’s your AI strategy?

It’s almost an understatement to say that “artificial intelligence”, or “AI”, is a hot conversation topic right now. We’ve put it in quotes because there’s vigorous debate over how “intelligent” the technology really is, but like it or not, the name “AI” is here to stay. There’s a lot of hand-wringing and fantasizing about what the technology could do for (or to) commerce and society, and business leaders are now regularly asked, “What are you going to do about it? How will you react to this new technology?”

We’ve been here before

Before we answer that question for Rhyze, let’s take a step back and do some due diligence on the technology itself. It’s helpful to note that although we’re in the throes of an intense hype cycle right now, we’ve been here before. To some, AI seems to have burst on the scene over the past couple years, but the term was coined in 1955, and the perceptron, a primitive but foundational form of the technology in use today, came just a couple years later.

Without getting too hung up on the details, we’ll note that these early developments sparked a lot of excitement and ambition—Marvin Minksy, one of the field’s early pioneers, was quoted in Life magazine as saying, “In from three to eight years we will have a machine with the general intelligence of an average human being.” This was in 1970; the hype started to die down when claims like that didn’t exactly pan out.

Fast-forward a few decades, and the technology has advanced by leaps and bounds, but we’re still working on that whole “intelligence of an average human being” thing (and ideally, we’d prefer an above-average intelligence, wouldn’t we?).

Two types of AI

To properly discuss our response to AI, it’s important to define some key terminology. AI systems can be divided into two main categories: they’re either predictive or generative.

An easy way to think of predictive AI is as a supercharged statistical model. These are the kinds of models businesses have used for generations to quantify risk, allocate resources, and so on. They’re simply a whole lot faster now and capable of digesting orders of magnitude more data to make a decision.

Of course, these models’ number-crunching prowess doesn’t mean they’re infallible. Due diligence is still required on both the data used to “train” the model and the results; otherwise, you end up with a model that predicts whether a patient is lying down when you think it’s predicting whether they’re ill.

At Rhyze, due diligence is one of our foundational values. As we scale quickly, we also want to scale correctly. We’re laser-focused on identifying and securing great investments for our customers and don’t want to distract ourselves double-checking predictive models and the data used to train them. We’re committed to investing in research, including the best ways to integrate this technology in the future. For now, though, rather than risking the distractions and inaccuracies of delegating crucial decisions to a statistical model, we’re leaning on tried and true methods we’ve developed over decades of experience.

So that’s predictive AI, the modern version of a longtime enterprise workhorse. The AI grabbing all the headlines right now, both for its uncanny ability to mimic human communication and for the astounding volume of copyrighted data used to train it, is generative AI. In a sense, generative AI models are also statistical models, but as the name implies, their focus is on generating new data rather than classifying their inputs in some way.

What can this type of AI do for us? Well, let’s pose a relevant question to the chatbot of the hour, ChatGPT:

Us: Can you describe in a few sentences why my credit union should invest in commercial real estate?

ChatGPT: Investing in commercial real estate can offer several benefits for your credit union. Firstly, it provides a diversified revenue stream beyond traditional banking services, potentially boosting profitability. Additionally, owning commercial properties can serve as a hedge against inflation, as real estate tends to appreciate over time. Furthermore, by investing in commercial real estate, your credit union can contribute to community development, supporting local businesses and stimulating economic growth while potentially enhancing the institution’s overall asset portfolio.

OK — not great, but not terrible. That’s not really the information we were after, though, so let’s see if changing the wording of the question changes ChatGPT’s opinions:

Us: Can you describe in a few sentences why my credit union should incorporate commercial real estate loans into its loan portfolio?

ChatGPT: Incorporating commercial real estate loans into your credit union’s portfolio can diversify revenue streams and mitigate risk. These loans typically offer higher yields compared to traditional consumer loans, potentially boosting profitability. Moreover, by serving businesses in the community with their real estate needs, your credit union can deepen relationships, attract new members, and foster local economic growth. However, it’s essential to conduct thorough risk assessments and ensure your credit union has the expertise and resources to manage commercial real estate lending effectively.

Those both look plausible enough as answers — they’re grammatical, at least. But note a few things:

A brief introduction to prompt engineering

The last point — that ChatGPT is highly sensitive to the wording of the question it’s given — deserves a little more exploration. The fact that a slight change to the wording can alter the context and information content of an answer can be risky if you ask it, for example, a detailed question about financial regulations.

The fickle art of getting the best results from generative AI by massaging the wording of your question (or “prompt”) is commonly known as “prompt engineering”. Put simply, in order to get the most accurate answers from a generative model, it’s best to:

In other words, it helps to have a decent idea what the answer should be before you ask the question. If that sounds a little backwards, we understand. It does underscore, though, that you can’t take text that a model generates at face value; it’s going to be influenced both by the data the model saw during training and the wording of the question it’s answering or text it’s responding to.

You can get grammatical yet ultimately useless answers from innumerable websites; you don’t need them as investment advice. Besides, as a customer-focused, service-oriented CUSO, we’re looking to build a long-term relationship with you, not cut corners by shuttling you off to a chatbot.

Still, even generative AI has its place. An internal rollout, where the technology is phased in gradually, can improve efficiency, as our friends at TruStone Financial Credit Union have found. It can work well as an aid for drafting documents, summarizing articles or policies, answering procedural questions, and similar constrained tasks. We occasionally use AI internally for such purposes, but a human always reviews and revises its output.

The bottom line

Both predictive and generative AI have their place, but they’ll always require human oversight. Use of the technology itself can create a number of additional risks, as discussed in a recent report by the US Treasury. We’re not going to delegate our core business functions, which will always be relational and experience-driven, to the technology of the moment. This hype cycle isn’t the first of its kind, so we’re going to stay the course and keep doing what we do best: thoughtfully evaluate investment opportunities and build lasting relationships with our valued partners and customers, while keeping an eye on the ever-evolving world of technology.

Partners

Everwise Credit Union is Indiana’s largest Credit Union with more than $5 billion in assets, over 50 branches throughout Indiana and southwest Michigan, and more than 300,000 members. Everwise offers traditional financial services including checking, savings, mortgages, and credit cards, as well as non-traditional services such as investments. Since its founding in 1931, Everwise has focused on making a positive impact on its members and the community. For more information, visit everwisecu.com.
Allied Solutions is one of the largest providers of insurance, lending, risk management, and data-driven solutions to financial institutions. Allied Solutions uses technology-based solutions customized to meet the needs of 6,000 clients along with a portfolio of innovative products and services from a wide variety of providers. Allied Solutions is headquartered in Carmel, Indiana, with regional offices in Florida, Minnesota, South Dakota, and Texas. Allied Solutions is a wholly owned and independently operated subsidiary of Securian Financial Group. Visit alliedsolutions.net to learn more.
Striving to provide its members with the personalized service, knowledge and honesty they deserve in a financial partner, Vantage West is proud to be among the largest credit unions in Arizona. From a one-room Air Force barrack over 60 years ago to serving nearly 200,000 members as the largest credit union in Southern Arizona, Vantage West has been dedicated to earning and rewarding the loyalty of the people and communities they are connected with, so they can all thrive together. Visit vantagewest.org to learn more.

Who is Rhyze?

Rhyze Solutions, a Credit Union Service Organization (CUSO), is a capital markets company that utilizes loan products, services, and technology to enhance the sources and uses of liquidity to its network of financial institution partners.
1313 N. Market Street
Suite 1000
Wilmington, DE 19801
info@rhyze.com

1-877-RHYZE11

switch to dark mode   switch to light mode