Competition for deposits from giant institutions on the one hand and aggressive fintechs on the other has put traditional community-based financial institutions on the defensive. Both the large banks and the startup players have been able to capitalize on highly targeted campaigns guided by artificial intelligence tools.
These AI-powered efforts can drive positive results without necessarily requiring the payment of high rates. Anyone can pay up for deposits, but such money rarely sticks absent some other reason or relationship. And the expense is usually not sustainable for institutions with high-cost branch networks.
Community banks and credit unions can now take a page from fintechs’ playbook and improve the efficiency and cost-effectiveness of their deposit acquisition and retention efforts. Most fintechs don’t build AI models themselves, observes Keith Henkel, founder and CEO of FI Works. They typically buy applications that have built-in AI modeling processes. Community financial institutions can do the same.
Henkel cautions, however, not to get the idea that banks and credit unions can just go buy AI off the shelf and start applying it. It takes special skill sets, technology and data preparation to make it work, even when it’s embedded in CRM or other marketing applications. Nor is it likely to work for institutions that have too few customer interactions to properly populate the models. “The more data you have the better off you are in the artificial intelligence game,” Henkel said during a webinar sponsored by The Financial Brand.
With those realities noted, community banks and credit unions really don’t have much choice but to look into greater use of AI. Deposits are the lifeblood of banking institutions and the cost of acquisition has shot up in recent years, with institutions bidding for deposits with incentives of $200 to $300 or even higher depending on the product.
Despite such incentives, there’s no guarantee that the institution is acquiring the right people, Henkel observes. The average new-account attrition rate after six to 12 months is 28%, he says, and can be as high as 35%. Further, 65% of acquired deposit customers turn out to be single-service households, on average. “All these things can be attacked with a good strategic plan supported by analytics and artificial intelligence,” Henkel states.
Machine Learning is the AI Sweet Spot for Most Financial Institutions
Before laying out several specific examples of how AI can help community financial institutions, Henkel addressed the confusion that makes the subject difficult for many banking executives.
AI is an umbrella term that can mean many different things, he says. People often associate AI with smart speakers like Amazon’s Echo or Google Home, or self-driving cars. These are in fact sophisticated uses of AI, but at the other end of the spectrum are data analytics and predictive analytics, which are really just the application of math and statistics long used within banking by credit card issuers and large financial institutions.
“Overall,” says Henkel, “think of artificial intelligence as a set of technologies that learns patterns from data.” Simple segmentation software, predictive analytics, data mining, robotic process automation, machine learning, and deep learning all are terms which fall under the AI umbrella.
” By using AI model scores instead of classic age-based segmentation, banks and credit unions are using science to figure out what somebody might need versus just throwing out offers.”
— Keith Henkel, FI Works
Machine learning (ML), which uses advanced algorithms combined with increased computing power, is the most relevant for banking applications at present, Henkel says. With ML being incorporated into CRM applications, banks and credit unions can begin targeting individuals based on AI model scores instead of classic age-based segmentation.
“Basically,” says Henkel, “you’re using science to figure out what somebody might need versus just throwing out offers, which too often is what happens.”
- Success With AI in Banking Hinges on the Human Factor
- 6 Fresh Ideas to Build Deposits When Other Banks Turn Up The Heat
- Building an AI-Powered Financial Institution
How To Put AI to Work at Small and Midsize Institutions
Success with AI models in large measure depends on asking the right question, and asking it in a form the model can answer, Henkel explains. The process begins with a retail or marketing executive who understands the business problem. This could simply be: How do we cross-sell more deposit products to customers who only have a CD?
From there, the institution needs a person with the knowledge and skills to use AI models, to begin to create more specific questions to be modeled and assemble the necessary data to feed into the model. If a bank or credit union wants to find out what products and services its customers are most likely to buy, says Henkel, they can build a model around their existing customers by asking the question: “What household is likely to add product X over the next six months based on customers that added it in the last six months?”
“What you’re doing basically is trying to match the segment of customers that don’t have the product with those who added it recently,” says Henkel. That process is often called “lookalike modeling,” he notes. At the end of the process, financial institutions end up with a score, which is applied to each individual whose data is in the model. Next step is to think about how to deploy that information — how to target specific households for marketing actions.
Good Data is The Lifeblood of Artificial Intelligence
All this requires a bank or credit union to have two things: 1. the right skill sets to not only set up the right questions for the model, but deploy the results, and 2. really great data.
“80% to 90% of the effort in using AI like this is getting the data together,” observes Henkel. That’s a challenge, he adds, because all financial institutions have different silos of data and different core system components, all hindering effective use of data. In addition, it’s often necessary to enhance the data you have in-house with outside data, particularly in regard to acquisition campaigns among prospects.
Finally, says Henkel, a trained human eye needs to review the data to be sure a model can make sense of it — looking for anything that doesn’t seem right.
” Clean data is essential. Don’t even start if you don’t have it. This is where many AI projects fail.”
— Keith Henkel, FI Works
“Clean data is essential,” Henkel states. “Don’t even start if you don’t have it. This is where many artificial intelligence projects fail. For many traditional financial institutions beyond the largest, he continues, “the work of pulling the data, consolidating it and prepping it for AI use is probably best left to experts, such as a marketing partner that has the technology and is already doing this work.”
- AI Advantages in Banking Grow, Adding Pressure for Broad Adoption
- Building a Deposit Machine that Won’t Cannibalize Existing Accounts
- Is This Community Bank’s Bold Digital Play The Model Of The Future
Examples of Acquisition, Cross-Sell and Retention
If all that seems like a great deal of work, remember that large banks and fintechs are already well ahead of most community financial institutions in using AI, which will only increase their competitive edge.
But equally important, the use of machine learning scores for customer acquisition, cross-selling and retention can pay off handsomely. In one case that Henkel cited, a $4 billion bank used an AI-generated score as the basis for offering different product offerings to different households. The bank used direct mail, but it also had individual branch bankers call people who had higher scores to try and close the deal. The campaign resulted in $25 million in new deposits, according to Henkel.
For converting prospects to new customers, the same bank used AI to help it identify the key characteristics of its best customers and then build scores among prospects indicating similar characteristics. Here the institution had to make use of third-party data for the prospects to score them. This campaign resulted in $50 million in new CDs and money market accounts.
Improving customer retention is a more difficult analytical challenge because a bank or credit union is studying historical transaction data, Henkel notes. One institution he worked with reviewed closed checking accounts, noting which ones had been getting regular deposits, and which ones showed regular deposits stopping. The idea was to be able to predict active accounts at risk of closing.
Timing is everything, Henkel notes. “You’re looking for people with scores at a point where you can still get in front of their decision to leave.” In the case he described, the institution saw about a 23% drop in attrition by using AI-based scoring.
All this must be “baked into your marketing processes,” for AI to be successful, Henkel states. Each institution must figure out how to get the scores in the hands of its sales staff so that when a customer walks into a branch or calls on the phone or chats, the retail banker is armed with answers and data to help convert a contact into more than just a question answered.