The Bilingual AI Banker: How JPMorgan Is Quietly Replacing Translators with Algorithms

On a weekday morning, you stroll through the lobby of JPMorgan’s Madison Avenue headquarters and notice something odd. The old, familiar sound of the bank floor, including phones, footsteps, and the quiet murmur of traders speaking three languages simultaneously, still permeates the building.

However, something more subdued is taking place upstairs on the floors where junior analysts and compliance officers used to stoop over bulky stacks of multilingual contracts. The number of translators is declining. Not all at once. Just slowly, almost courteously.

Given that JPMorgan, of all banks, is not known for making abrupt changes, it is difficult to ignore this change. For years, Jamie Dimon has stated that the company prefers to implement new technology in a methodical and controlled manner. However, between 2023 and the present, the bank’s in-house LLM Suite has started to perform tasks that were previously performed by qualified human linguists, such as translating loan agreements, summarizing Mandarin earnings calls, and writing client memos in German that previously required a specialist with a background in finance and a passport’s worth of stamps.

Speaking with people in the financial technology community, it seems like JPMorgan didn’t actually make this announcement. It simply occurred. With over $4.4 trillion in assets and an annual technology budget of more than $18 billion, the bank has been subtly integrating multilingual AI into almost every back-office niche. Generative AI has the potential to unlock $200 to $340 billion in annual value across global banking, according to a McKinsey estimate. As the top-ranked company on the Evident AI Index, JPMorgan appears committed to capturing as much of that as it can.

Launched in 2024, the LLM Suite is model-agnostic and draws from the bank’s internal data layers, OpenAI, and Anthropic. The platform now completes tasks that used to take hours for a junior analyst in about thirty seconds. It was asked to translate and summarize a Cantonese earnings call by a Hong Kong investment banker. According to multiple internal accounts, the outcome was identical to the work of a senior translator. Perhaps better. Perhaps a little soulless. Which is still unknown.

Employees at the bank report increases in productivity of 30 to 40 percent. In less than eight months, two-thirds of the staff have been added to the platform—a rapid pace for a company known for its caution. The roles of contract reviewers, translators, and some compliance specialists have become more limited. For AI oversight work, some have received retraining. Others have just moved on, silently wondering what happens to language experts when their specialty is incorporated into a chatbot window.

It’s possible that the bank is correct and that this is simply the next logical step, similar to how ATMs used to replace tellers without putting an end to the industry. Investors appear to think so. The annual value of JPMorgan’s AI initiatives is estimated to be around $1.5 billion, and the figure continues to rise. However, there’s also a sense that something subtle is being lost as you pass those desks that seem more empty. Tone is captured by a human translator. Grammar is caught by a model. Even if the output appears the same on paper, the two are not the same.

Years ago, both Tesla and the first online brokerages encountered similar skepticism. The pattern is recognizable. The world gradually adapts after the technology is introduced and the skeptics scowl. It’s unclear if the bilingual AI banker at JPMorgan will become the new standard or the next cautionary tale. However, the algorithms are already communicating. The bank is listening for the time being.