U.S. Bank bets AI can finally deliver 360-degree view

U.S. Bank receives thousands of customer calls a day. But this July 4, it could tell for the first time which ones were military service members or veterans: 450 in all.

“We were able to say, thank you for your service,” said Bill Hoffman, chief analytics officer at U.S. Bank.

What’s noteworthy about this is the way the service members were spotted: by Einstein, Salesforce’s new artificial intelligence component to its customer relationship management software. The bank has been piloting Einstein for a few months.

Such small nuggets of knowledge can be surprisingly powerful, said Hoffman, who also heads the bank’s CRM.

“This doesn’t always have to be about knowing a customer well enough to provide a new mortgage offer or talk about financial planning,” he said. “It could simply be, ‘Happy birthday, thank you for your service, here’s where you’re at with your mortgage application.'"

Like many banks, U.S. Bank is seeking a real-time, 360-degree view of its customers, so that when clients call or walk into a branch, the rep knows about all the products and services they use with the company, their history, their background and all recent activity. For instance, what they’ve done lately on the bank’s website and mobile app or any calls with questions or complaints. Armed with this knowledge, the rep could complete a transaction that had been broken on the website or suggest a useful product or portfolio adjustment.

Bill Hoffman, chief analytics officer at U.S. Bank.

Companies once thought customer relationship management software would accomplish this, then Big Data and analytics. But the 360-degree view of the customer has eluded most financial services companies largely because of their data silos — typically, bank account data is in one place, mortgage data is in a separate system, the contact center has its own platform and so on. There’s no easy way to aggregate all of these and assess them in real time.

With its partnership with Salesforce, U.S. Bank is signaling its hope that artificial intelligence will do the trick. The company’s new Einstein software applies machine learning, deep learning, predictive analytics, natural language processing and data discovery to customer data, and is designed to be used to predict future behavior and recommend best next actions.

Other banks are betting that AI help employees better interact with customers. Morgan Stanley is using machine learning to give financial advisers “next best offer” suggestions. DBS Bank has a chatbot that can perceive when customers are frustrated with automated answers and hand an interaction, along with recommended next steps, to a human. Others, such as Bank of America with erica and Capital One with Eno, are applying AI to automated chatbots that will be designed to understand and predict customer needs.

“Banks have been struggling for years to effectively use the increasing amount of customer data they hold,” said Jim Perry, senior strategist at Market Insights. “AI enables new insights about your customers that help you be of better service to them. We may feel we know our customers, but AI and machine learning have the capacity to augment our knowledge.”

Michael Abbott, digital lead for Accenture financial services, North America, said that for the past three years, large financial institutions have been doing “science fair projects” around customer intelligence.

“It’s been bespoke efforts, like building a car in the 1800s, where you need to pull together disparate pieces of data and organize them,” Abbott said. “Then you need to set up a whole bunch of hypotheses around what you want to find out, then you need to take wicked-smart people on the decision-science side and test those hypotheses.”

Now, banks need to transition these projects into industrial tools that systematically extract the signals from the noise and use those signals across channels, he said.

A signal might be someone who’s had a dramatic increase in debit use over the last several months in an area far from where they live. The person might be looking to move, and therefore the mortgage group might want to know about it.

“AI is a tremendous way to capture multiple patterns to find a signal,” Abbott said. “Think of a scene, where you have trees, a house, someone mowing a lawn, a no-parking sign, a stop sign down the street. What AI is fantastic at doing is looking at that whole entire scene and saying I want to pick out the stop sign. You could never pick out the stop sign using traditional techniques.”

Giving AI to the front line

Hoffman has spent his career, which has included stints at the Central Intelligence Agency and several consulting companies, leading analytics teams to help executives and business line leaders to make better decisions. At U.S. Bank the focus, at least for now, is on the customer-facing staff.

“We feel passionately that enabling good decision-making at the front line, in the moments that matter with our customers, is a game changer,” he said.

AI is needed here, he said, because it’s not possible for human beings to absorb and interpret all the data quickly enough. “I think there’s an opportunity for us to leverage AI and machine learning to make sure we have a real-time or near-real-time understanding of how our customers are interacting with us,” Hoffman said.

U.S. Bank is applying Einstein to more complex relationships in commercial banking, wealth management, small business and mortgage. The bank partners with another vendor, which Hoffman declined to name, for higher transaction volume, less complex relationships.

The AI engine can suggest a next best action more accurately than a plain analytics engine, Hoffman said. Analytics might suggest the best conversation to have next with a customer, based on past transactions, would be about a new credit card offer. But the AI software could know better: that the customer had a few hours earlier been on the bank’s website, finding out about a mortgage.

Einstein might be able to pick up on predictive patterns, too. It might discover that a customer walking into a branch and doing two or three things indicates something.

“Einstein is allowing us to see a more holistic view of the way our clients are interacting with us,” Hoffman said. “As a result, some of the nonobvious indicators can come to the fore. That's what we're doing right now with Salesforce, combing through and understanding, are there some nonobvious indicators or signals that matter that we can leverage and surface for a deeper, better client relationship?”

Einstein could alert a rep that the customer they are interacting with called the contact center the night before and provide insights to the mood of that call. It could spot signals of customers looking to refinance, giving the bank a chance to incentivize them to stay. It could alert small-business lenders that an entrepreneur recently contacted a U.S. Bank wealth adviser, giving them an opportunity to cross-sell that customer.

Ultimately, these examples are some of the ways that AI could help the bank provide value to their customers the same way that Amazon suggests products or services.

“We know that the bar is only going higher with some of the best-practice providers across the industry, inside or outside of financial services. The expectation has been that customers want us to know them, to recognize them and to understand them while they continue to grow.”

Laying a data foundation

To get all this to work, U.S. Bank had to do some data integration.

“You can’t just build out a new CRM system,” Hoffman said. “You have to do your back-end work to integrate the data.”

Divisions like the mortgage company or small-business banking have a strong understanding of their customers, but haven’t historically had a good sense of customers’ total relationship with the bank at large, Hoffman said.

“In many cases, the power plants are up, but the lines are down — you have pockets of great data on a customer, but folks have not spent the time to unify the customer experience by integrating that into a holistic view.”

With the integration work done, Einstein can draw on data from different business lines.

The path forward

The bank plans to test Einstein for at least one more quarter. If all goes according to plan, it intends to expand its use to other parts of the bank in 2018 and beyond. This is part of a broader implementation of CRM that’s been under way since 2015 and will continue through next year.

Hoffman sees the work with Einstein as a learning experience.

“We have good ideas, but we don't have a monopoly on them within the four walls of U.S. Bank,” he said. “So we want to partner with leaders like Salesforce to learn, to make sure we’re leveraging best practices from financial services and other industries. We are spending a lot of time very intentionally in that pilot mode — we don't have this all figured out, we want to make sure we're using this as a mechanism to learn.”

The future goal will be to help the bank be predictive in anticipating their needs.
Abbott says U.S. Bank is in the vanguard, and other banks are sure to follow.

"You’ll see sophistication grow exponentially over the next several years," Abbott said, "and it will be refined as people learn more and learn how to multiply these techniques and learn how to deploy them through product development and channel and cross-sell strategies.”

Editor at Large Penny Crosman welcomes feedback at penny.crosman@sourcemedia.com.

For reprint and licensing requests for this article, click here.
Artificial intelligence Machine learning Customer data CRM systems U.S. Bank BankAI Conference
MORE FROM NATIONAL MORTGAGE NEWS