Can AI make banks as good as Amazon at knowing customers?

Register now

When you buy a book from Amazon, you know you’ll get several book recommendations based on that purchase and other past purchases.

The suggestions won’t be about what other people in your age group have bought or what people in your neighborhood liked. And they won’t be based on months-old data.

They’ll be very specific to you.

[Digital identity is broken, and fixes are urgently needed. Learn how large financial service and health care companies are tackling the issue – to enhance customer experience, to stake out positions in their business ecosystems, and to manage risk – on our Feb. 23 web seminar. Click here for details.]

When you make a bank transaction, you’re unlikely to receive any recommendation or advice, even though the bank may have more relevant data about your money than Amazon does.

Banks struggle with customer analytics. They have a lot of data, but they also have old databases, fiefdoms that don’t want to share customer data with one another and a problem in identifying the best ways to use the data.

If banks want to get serious about providing the same type of personalized insight that big tech companies do, they’ll need to get serious about adopting artificial intelligence first.

“The Googles, Amazons, Facebooks and Apples have set the standard,” said Stephanie Sadowski, managing director at Accenture. “The moment banks start to leverage this data to deepen the relationship and personalize it, we’ll start to see them use AI in a push fashion” the way Amazon does.

For example, through AI-powered analytics, a bank may notice that customers could get a higher return on their money by moving some cash to a different account. The fintech startup offers suggestions of this nature, as do virtual assistants from companies like Personetics, Kasisto and North Side.

“It’s going to be a customer expectation that you tailor your conversation to my needs and my preferences,” said Alvi Abuaf, senior vice president and head of financial services consulting at Capgemini Consulting. “When customers interact with Google, Amazon or other players like that, they know what they’re being offered is very much customized to them.”

The mandate
Some argue that banks have to get better at customer analytics to stay relevant, compete and grow.

“There’s no question that customer analytics and specifically AI for marketing and cross-selling purposes will be a key success factor for banks going forward,” Abuaf said.

Intense competition from fintech companies is one driver, especially as those fintechs increasingly get access to banks’ data through aggregators and open APIs.

And Europe’s Directive on Payment Services (PSD2), which forces banks to share their customer data with others, may start to have an impact across the pond.

“Will it come here as a regulation? I don’t know,” Abuaf said. “Will it come here as a market-driven direction? Absolutely.”

The challenge
Banks often get criticized for not being good at customer analytics. But observers say that might be a function of priorities.

“Over the last 10 years or so, banks have accumulated a lot of data and they haven’t figured out what to do with it yet,” Abuaf said. “That’s not because they’re not capable, but mainly because they’ve been focused on regulatory issues.”

Capgemini’s view is that that will change in the next year as banks start focusing on growth, new products and new services. As they do, they’ll recognize the value they have in that data.

Banks currently tend to use simulations, customer segmentation and life-stage moments, said Luc Burgelman, CEO of the customer analytics software company NGDATA.

Such older techniques don’t achieve that coveted individualized “segment of one” insight.

Google and Facebook don’t use these older techniques. “Why simulate reality when you can look at reality?” Burgelman said.

Another specific challenge for financial companies is that their data is temporal, noted Scott Howser, CEO of Nutonian, another provider of AI-driven customer analytics software. “Every customer experience and market condition is changing with time,” he said.

Data analytics projects tend to take too long to keep up with this fast-moving data, Howser suggested.

“A dozen Ph.D.s sit in a room and build models and translate them to something the business can take action on,” he said. “That entire life cycle is far too long, it doesn’t meet the needs of the business or the consumer.”

Banks also have a basic data challenge: customer data tends to be locked in disparate databases based on products. And customers might be identified differently in each system. One might identify the customer by legal name, another might use Social Security number, a third might use a bank-generated customer ID number.

What AI can do
There are a few ways artificial intelligence technology can help banks overcome these shortcomings.

Some AI-driven programs, including those from NGDATA and DataXylo, are designed to take care of that customer identifier problem. They use fuzzy logic to match customer records despite the out-of-sync customer tags.

“We use a combination of deterministic and probabilistic approach, so if certain attributes like your date of birth are matching, those records could be clustered together,” explained Abhishek Yadav, co-founder and CEO of DataXylo. The software could also link several family members to one common household.

Also, because artificial intelligence software can continuously learn, it can better understand customers’ reactions and preferences over time than a human, who can’t remember and digest all that data for each customer.

AI can respond much faster to changes in customer behavior or markets than traditional customer analytics can, Howser said. It can analyze data in real time, which makes for a more precise recommendation.

“Every time there’s an interaction with a client, you get smarter about that client and your actions are going to be better,” Burgelman said. The artificial intelligence also fills in any blanks due to missing customer data, using techniques like propensity scores and collaborative filtering.

The system might wait until the moment a customer is in a car dealership or on a car sales website, then send that person a car loan promotion.

“It’s like being a sniper and waiting for the right moment,” Burgelman said.

Editor at Large Penny Crosman welcomes feedback at

For reprint and licensing requests for this article, click here.
Artificial intelligence Analytics Technology Predictive analytics