Bank of America CEO Emphasizes AI Data Accuracy

Brian Moynihan, CEO of Bank of America, told the Forbes Iconoclast Summit that customer-facing AI in banking must deliver precise answers, saying, "the data has to be perfect," per American Banker. American Banker reports Moynihan said the bank has spent roughly $250 million so far this year on new AI deployments, within a technology budget of about $13 billion a year that includes $4 billion for emerging technology such as AI. He cited Erica, the virtual assistant Bank of America launched in 2018 and developed with Stanford researchers, which American Banker reports is now used by about 20 million customers roughly 200 million times a quarter. Moynihan said that for anything requiring real judgment, a human has to remain in the loop, according to American Banker.

Editorial analysis: Customer-facing financial assistants demand deterministic accuracy, low-latency retrieval, and verifiable data provenance. Teams building comparable systems typically invest in end-to-end data quality, identity linking across accounts, reconciled transaction ledgers, and automated test suites to catch silent failures before they reach customers.

American Banker cites commentary from Theodora Lau of Unconventional Ventures and Mate Jendrolovics of Intuitech, who note that partial accuracy can be functionally worthless for banks. Editorial analysis: regulated institutions face asymmetric costs from model errors - mistakes can touch customer funds, compliance reporting, and trust - which raises demand for grounding, verification, monitoring, and explainability tooling across the vendor landscape.

Editorial analysis: Watch disclosed AI spending levels at large banks, any operational incidents tied to assistants like Erica, and vendor activity around data lineage, evaluation, and production monitoring - the fidelity concerns Moynihan emphasized.