‘AI isn’t just technology, it’s good judgment’: Cathy Bessant on why banks need humans

Despite prevailing wisdom, Cathy Bessant doesn’t want Bank of America to fail fast or often. That may be the mantra of tech companies in California, and banks largely see themselves as tech companies, but that’s actually where they differ. “We have a huge responsibility, we happen to love that and think it’s our job to steward it well,” said the bank’s chief operations and technology officer. “The safety and preservation of our customer trust means that lots of failures can not be in the cultural way we think.” Bessant is transforming B of A into a technology firm and leading it into a future that requires humans to apply immense computing power to immense amounts of data — an artificial intelligence future. But like any every other technology executive, she’s facing a shortage of talent with science, technology, engineering and math backgrounds. The result is often departments hiring from other departments. Bank of America will spend $600 million this year on cyber defense alone. It employs 1,200 people whose jobs are dedicated to nothing else but information security, although the company makes that “the job of every single employee,” Bessant said. But the caliber of the talent the bank hires is just as important as the technology it invests in, since they'll be the ones that determine how to work responsibly with AI. “Sometime’s theres a misunderstanding that cyber defense is somehow all technology, and it’s not," she said. "It’s judgment and how to assess a threat and think about mitigation every single day.” Banks have always served first and foremost as guardians of people’s money. Today that extends beyond dollars and cents to credit card and Social Security numbers and other personal and trackable data created on consumers and by consumers at a rate that’s now doubling every two years. In order to maintain customer trust in a technology-driven world, the humans of Bank of America need to make sure they’re using data responsibly, knowing that by serving their customers well they may end up using data in ways customers never anticipate or intend. For example, if a customer uses a B of A credit card to buy an airline ticket overseas, the bank has the potential to not require him or her to set a travel indicator, Bessant explained. “We know you used your credit card properly to buy your airline ticket and we know when you’re there, and if you used your credit card, why should you have to tell me?” The holy grail of AI is to have employees responsibly apply the technology, the computing power, to data, the models and algorithms themselves and the workforce — and to know when to apply it to which, Bessant said, using a driverless car as an example. “In the midwest, it doesn’t matter what the driver’s training manual says, we are taught to hit the deer,” said Bessant, a Jackson, Michigan native. “If a deer jumps out onto the road, you don’t go off in the ditch because you don’t know what's there; you don’t go into the other lane because there could be oncoming traffic… the safest and best decision you can make is to hit the deer.” Diversity of talent is important. When hiring the person that’s going to write the equivalent algorithm for a driverless car, bank leaders need to consider where they learned to drive, how they assess what’s in front of them and what actions they can and should take. Bank of America patches 40 million vulnerabilities a year, which is 20 times its rate from five years ago. That’s 40 million times a human needs to decide what to patch in what order, where in the system has the biggest risk of instability and how to balance the pace of change with the patching. “The decisions that will make your driverless car work everyday will be made by people you will never see.” Technology is the new competitive advantage — not just for banks for all non-designated tech companies that are, like everyone around them, being driven forward by technology. Bank of America currently has 81 patents or patent applications related to AI to protect its competitive advantage; Bessant counts it as proof to technologists that it is a serious technology company (and doesn’t have to fail fast, often and everywhere to be one) that’s innovating aggressively and hiring the highest caliber of people — globally. “Not all good wine comes from California, and not all good technologists are in California,” she said. "You can create IP no matter where you are geographically in our firm. It’s been a big selling point on campuses because it's a proof point around our culture of innovation." AI won’t ruin the workforce or decimate the need for humans, she says, “unless we’re cavalier about it and don’t pay attention as stewards of employees… If it does, shame on us.” It is forcing organizations to reskill the workforce to different things that will be required in an AI world. For example, rather than having number crunchers process data, a company should reskill those people to write the model that produces the same results, test and validate those models and keep them healthy as time passes and the environment changes. During the mortgage servicing crisis Bank of America had hundreds of people who worked with customers restructuring their mortgages, Bessant said. That portfolio has been reduced dramatically and in the last five years the bank has re-deployed and re-trained between 3,000 and 3,500 people. “You can automate a task but if regulation changes you have to reprogram the way you automate,” Bessant said. "You’re not counting the money but you’re keeping the robot healthy.”