Bank of America says its AI will return $700 million next year. That is under 1% of its costs, and your CFO now has a benchmark

Bank of America put a dollar figure on AI on its third quarter earnings call on October 14. CEO Brian Moynihan said the bank has about 130 AI projects in production, that all of them pay back, and that he expects them to return two to three times their cost over time. He put the benefit at roughly $250 million to $300 million this year, rising to $700 million next year, per the call transcript(opens in a new tab).

That is the number to study, and mostly for its size. On the same call, the bank reported adjusted expense of a little more than $18.5 billion for the quarter, up almost 7% from a year earlier. At that rate the cost base runs near $74 billion a year, so $700 million is under 1% of it. Set it against the $4 billion a year Moynihan said the bank spends on technology change, and it is still just under a fifth. One of the most AI-active banks in the world is booking real money, and the money is modest.

The headcount line tells the same story. The bank ended the quarter with 209,911 employees, down nearly 3,500 from a year earlier, which CFO Alastair Borthwick tied to automating processes and simplifying work. Managers fell by about 5% while client-facing staff grew. That is a slow, managed shift, not a wave of AI layoffs.

The contrast with JPMorgan, which said a day earlier that AI benefits were not yet visible in its quarterly results, is mostly about disclosure. Bank of America chose to count. Once one large peer reports an annual AI benefit in dollars, a board will ask every other one for its figure.

The practical step is to build that figure before you are asked. Count benefit per project, in dollars, against a named cost line. Report it as a share of total operating expense, not as hours saved. And set the expectation now: a credible first year number may look like Bank of America's, real and small.