- Key insight: Bank of America is directly connecting increased revenue and decreased cost to AI.
- What’s at stake: Operational costs and competitive advantage hinge on successful AI integration.
- Supporting data: Erica has processed three billion customer interactions, replacing the equivalent of 11,000 staff.
Source: Bullets generated by AI with editorial review
Bank of America’s Erica virtual assistant is now doing the work of 11,000 call center reps. This is one way the bank is seeing measurable revenue increases and expense reductions as a result of its deployments of artificial intelligence, Hari Gopalkrishnan, chief technology and information officer, said at an event on Monday.
Bank of America has a $13 billion technology budget, $4 billion of which goes to new projects. The bank’s technology priorities are improving customer experience, driving revenue up and taking expenses down, Gopalkrishnan told Lananh Nguyen, U.S. finance editor at Reuters, during the Reuters AI Momentum conference in New York.
His remarks came at a time when many businesses are still waiting to see a payback on AI. An August MIT study found that 95% of companies are failing to get a return on AI investments.
But at Bank of America, “covering the eight different lines of businesses, there’s revenue opportunities everywhere,” Gopalkrishnan said.
AI is making relationship bankers more productive, for instance.
“Think about how much time and effort people take to prepare for client meetings,” Gopalkrishnan said. “How many clients can a given relationship banker cover if you’re able to actually automate that process using AI and traditional analytics and traditional automation? You now can have a banker cover 50 clients instead of 15, and that’s exactly what we’re seeing in the real world.”
These bankers use AI-based client meeting preparation tools that let bankers say, “I’m going to meet client X tomorrow. Can you prepare for me everything that the client cares about?” Gopalkrishnan said. “It can bring together structured data, unstructured data, everything I’ve talked to the client about from a customer relationship management perspective, and that’s at my fingertip with insights.”
On the expense-cutting side, AI helps the bank understand where it’s spending money, he said.
“Every second of handle time in a call center is worth X dollars,” Gopalkrishan said. “So if I can automate prediction of why a customer is calling, I can then look at speech to text to figure out what the client call need is, and give the associate an agent to serve them within 30 seconds or one minute. That’s 30 seconds of money saved every single call.”
Bank of America’s 8-year-old AI-based virtual assistant, Erica, has handled 3 billion customer interactions.
“We’ve already seen the equivalent of 11,000 people’s work being done every day through Erica,” Gopalkrishnan said.
In the past, if a customer saw a charge go through three times or wanted to order a checkbook, they would call the call center. “Those calls no longer have to be answered by people, because Erica takes care of them,” Gopalkrishnan said.
AI has been used to analyze fraud in person-to-person payments 7 million times, Gopalkrishnan said.
“Every time you make Zelle payments, we run two AI-based models to figure out if it’s you, if it’s the right amount, are you sending it to the right people?” Gopalkrishnan said. “That’s all automated through AI that’s been in production for a couple of years and is working, and we have significantly reduced our fraud losses.”
To protect against the danger of hallucination and error, Bank of America has a process by which it looks at 16 different dimensions of risk in these models, including hallucinations, ethics and privacy.
In wealth management, Bank of America has an AI-generated market analytics platform that sends all its financial advisors a synopsis of what happened in the market the previous day, which they can then overlay on their client portfolios.
“So you can call up your client and say, ‘I know you really are interested in what’s happening in Brazil. Here’s the three things you should know about that,'” Gopalkrishnan said. In the past, this would have required those advisors going to six different places, he said.
Overall, AI is taking “toil out of the system,” he said. “In three years, hopefully we will have significantly streamlined a lot of toil on the system and be doing more exciting work going forward.”