Bank of America plans to double its AI expense budget next year, with the expansion running toward 2027, even as chief technology and information officer Hari Gopalkrishnan cautioned that AI should not be the default answer. Speaking at Fortune's AIQ Summit in New York, Gopalkrishnan said the bank evaluates AI implementations across 16 risk dimensions, including privacy, bias, workforce implications and intellectual property. Deterministic models, mobile applications or real-time decision rules can be better suited to some tasks. The bank says about 140 AI and machine-learning projects cost around $400 million and generate an estimated $800 million benefit. Bank of America has about 200,000 active AI users, more than 400,000 daily prompts and over 300 approved use cases, while its Erica assistant has handled more than 3 billion client interactions. The bank reports 10% to 20% productivity gains among its 18,000 to 20,000 software developers and is relying on natural attrition rather than layoffs to manage headcount. Separately, S&P Global executives described a data and AI strategy centered on accuracy, citations, auditability and traceability. In work with an unnamed tier-one bank, S&P Global said it reduced time to market roughly sixfold and improved accuracy across multiple data sets from about 60% to 98%.