AI Gains for Big Banks Pose a Competition Headache

(Bloomberg Opinion) — Bank of America Corp. first launched its AI-powered chatbot Erica nearly a decade ago in 2016. The platform, which has been iterated several times and has received numerous patents, handles approximately 2 million customer interactions every day; This is equivalent to what 11,000 employees can do.

If that sounds impressive, the flip side is cost: The company spent nearly $120 billion on technology over the same period, and last year’s $12 billion technology budget included $4 billion for development, including improving Erica and developing new apps, on top of the $8 billion needed to maintain existing systems. These are huge sums of money, and investors at many major banks have long been asking what returns they would get on that cash. It’s a good thing that some answers are starting to emerge; But these are somewhat limited, and there are two important caveats to this story.

First, the costs are partly high because companies need to be extremely careful when deploying new tools, especially generative AI. Because mistakes can be devastating to trust and lead to wasted investment. Second, AI promises to increase competition issues because it looks set to put an even greater wedge between the biggest lenders who can spend the most and the rest of the pack.

Bank of America is an example: Its annual technology budget is larger than the entire cost base of more than half of the lenders in the KBW Banks index. JPMorgan Chase & Co.’s $18 billion in annual technology spending is more than all but five other banks in the index combined.

Details about BofA’s bang for its buck were the most interesting parts of last week’s investor day, its first since 2011. It was stated that the bank’s consumer arm reduced its headcount from 101,000 in 2011 to 55,000 this year, due entirely to better technology. He also added that it has halved fraud losses across the bank since 2018.

Artificial intelligence has been a big part of this. Instead of using Silicon Valley firms, BofA built everything itself, which made it a successor to Capital One Financial Inc. has made it one of the largest intellectual property owners in the field of finance. Wells Fargo & Co. According to analysts, these two account for 65% of all AI-related patents held by banks.

But while more and more companies are acknowledging their spending on technology, they still provide little insight into the true return on investment of AI, and available data is disappointing. Less than half of the 280 financial managers who participated in a survey conducted by Boston Consulting Group this year were able to measure the return on artificial intelligence investments. Of those who can, a third have so far pegged their repayments at less than 5 percent, while another quarter say it’s between 5 percent and 10 percent.

Part of the problem is that there’s no off-the-shelf product you can pick up and plug in, like there is with Microsoft’s Excel spreadsheets. Even those who choose to work with a large GenAI company, as Morgan Stanley did with OpenAI, still need to invest a lot of time and money to turn a large language model into a useful tool, whether it’s a public chatbot or an internal assistant for research or sales ideas.

Even before going this far, a firm must spend time and money on its data to make it useful for any AI project; This means cleaning, sorting and tagging all data. Morgan Stanley spent several years doing this before even starting to consider working with AI. Bank of America spent $3 billion between 2014 and 2019 to make its own data available. Banks do this for other regulatory and commercial reasons, but it highlights the costs of getting to the starting line for an AI project.

JPMorgan spends about $2 billion a year on artificial intelligence projects and last year announced that they generate nearly $2 billion in annual cost savings, most of it related to fraud. However, this does not mean that it will achieve a 100% return on investment; Many other data and technology spends have brought JPMorgan to the point where AI could start to be useful. Big tech budgets help big banks get ahead.

Even banks that can invest such mind-boggling amounts in software creation and development have to spend huge amounts of money on testing before launching products. BofA CEO Brian Moynihan made this point about the AI platform last week: “It has to be perfect.”

“If people lose confidence in that answer [from Erica]11,000 people need to be connected to phones and branches tomorrow. “Tomorrow,” he said emphatically.

This is not just about banks, whose duties to customers are strictly regulated. The basic dynamic Moynihan describes applies to every company, regardless of whether its AI users are individual customers, other companies, or its own employees. The ultimate rewards of AI in terms of efficiency and perhaps personalization of service may hold great promise, but the time and money required to achieve them are also large and often paid upfront. And there is no guarantee of success.

The more AI delivers on its promises, the more companies will become the ones that are already the biggest and richest. One day, perhaps sooner than we think, this will create a competition problem that politicians and regulators will need to start thinking about how to solve.