Bank of America plans to double its AI budget while warning against AI overuse

Bank of America plans to double its artificial intelligence budget next year. Its chief technology and information officer also thinks a lot of companies, his own included, reach for AI too quickly.

That sounds like a contradiction. It is closer to a spending philosophy: put more money into AI, and get pickier about where it goes.

The tech chief’s warning

Hari Gopalkrishnan, Bank of America’s CTIO, made the point at Fortune’s AIQ Summit in New York. He argued that AI often gets picked as the answer before anyone checks whether a simpler tool would work.

“One of the biggest mistakes we see us and others doing is rush to AI as a solution, when deterministic models do a plenty good job.”

A deterministic model follows fixed rules, so the same input always produces the same output. For a bank, that predictability matters. Regulators, auditors and customers tend to prefer systems that behave the same way every time.

The numbers behind the bet

Bank of America’s plan to double AI spending, with the expansion running toward 2027, is reportedly driven by strong returns on what it has already built. Employee-driven initiatives have also fed into the decision.

CEO Brian Moynihan has put figures on those returns. He said approximately 140 AI and machine-learning projects have gone live at a cost of around $400 million. The estimated benefit from those projects is $800 million, working out to roughly two dollars back for every dollar spent.

The AI push sits inside a much larger technology budget. Bank of America spends roughly $13–14 billion a year on technology overall, with around $4 billion of that going to new initiatives that are increasingly built with AI baked in.

How much the bank already uses AI

Adoption inside the bank is already broad. Bank of America counts about 200,000 active AI users who send more than 400,000 prompts every day across more than 300 approved use cases.

On the customer side, the bank’s Erica virtual assistant has logged more than 3 billion client interactions.

The jobs question

Bank of America’s approach so far has been to keep headcount stable through natural attrition rather than layoffs. The bank has reportedly seen productivity gains of 10–20% among its software developers, a group that numbers between 18,000 and 20,000 people.

What this means for banks and their customers

There are caveats worth watching. The $800 million figure is an estimate from the bank itself, and the developer productivity gains are reported rather than independently verified. Benefit estimates for internal technology projects can be generous. Time saved does not always turn into money saved, especially when headcount holds steady.