Why Wall Street is Using Bilingual AI to Decode Foreign Earnings Calls in Real-Time

Every foreign earnings call has a moment when an executive moves into a quieter register, typically around the fifteen-minute mark. The script that was prepared has been delivered. The CFO has rattled through the figures. And now a CEO from Frankfurt or Seoul is responding to questions from analysts in a way that seems a little strange. Not exactly dishonest. simply textured. layered. This type of language can cost an investor actual money because of the discrepancy between what is stated and what is intended.

For many years, it took a unique combination of native language proficiency, financial fluency, and the ability to spot deception disguised as accuracy to close that gap. It required assembling teams in Hong Kong just to listen or employing Mandarin-speaking analysts in New York. The majority of funds didn’t care. The trade had already progressed by the time they received the translated transcript.

That computation is changing quickly. Wall Street has discreetly implemented a new class of artificial intelligence (AI) that is multilingual, quick, and unsettlingly accurate. It can listen in on a live call in Korean or Japanese and identify the important signals in real time. Reading the spaces between words is just as important as translating words.

A look at what these systems are capable of is provided by research from Chicago Booth and Georgia State University. Researchers discovered that AI could identify subtle linguistic patterns predicting changes in corporate investment policy, such as changes in capital spending, dividend behavior, or even hiring plans, before they were formally announced by using ChatGPT to analyze nearly 75,000 earnings call transcripts from about 3,900 U.S. companies.

The scores produced by AI closely matched the information provided by CFOs in confidential surveys. That is a significant discovery. It’s a subtext-reading system.

The implications become more pronounced when you extend that capability across languages. When discussing margins, a CEO in Frankfurt who employs a particular cluster of hedging phrases is not speaking at random; rather, these phrases have meaning in the German financial communication tradition that a non-native ear might overlook. In addition to translating, bilingual AI interprets within linguistic and cultural contexts, highlighting the kinds of omissions and deflections that skilled human analysts must spend years learning to recognize. Businesses using these tools seem to be gaining something more akin to structural advantage than just a quicker workflow.

An additional layer is added by the vocal dimension. According to a January 2025 study in the Journal of Accounting Research, CEOs’ vocal patterns during earnings calls could be used by machine learning models to identify depressive symptoms. Over 14,500 recordings from S&P 500 companies over a ten-year period were examined by the researchers. Over 9,500 executives were evaluated. When you put it simply, it sounds almost intrusive, and perhaps it is, but the financial reasoning is compelling. A publicly traded company’s CEO in distress poses a significant risk. This has always been intuitively understood by the market. It now has a measuring tool.

The setup is still recognizable if you walk past any major trading desk in Manhattan: price feeds, candlestick charts, and Bloomberg terminals. However, the way intelligence interprets those charts has subtly evolved. Workflows that were previously entirely human, such as reading filings, identifying tone changes in transcripts, and summarizing Q&A sessions from calls that took place overnight in Asia, now use large language models. Analysts continue to be important, exist, and make decisions. However, the amount of data they are expected to process has increased beyond what is reasonably possible for human bandwidth.

It’s important to be objective about the problems that this doesn’t resolve. A human is still required to respond intelligently to an AI signal that indicates a change in policy or executive hesitancy. Sometimes subtleties that a native speaker would pick up on are flattened by translation, even excellent AI translation. What happens when several AI systems respond to the same signal at the same time, pushing prices before any human analyst has processed the call, is another serious question that remains unanswered. The trade-off between efficiency and volatility is not theoretical. Regulators are keeping an eye on things like this, even if they are unsure of how to react just yet.

The direction feels certain, though. Businesses that hired bilingual analysts for years to cover Asian and European markets now have a different kind of competitive resource: one that never sleeps, never misses a call, and learns from each transcript it processes. It’s difficult to ignore how subtly this change has occurred. No press conferences. No announcements. Just a gradual shift in the way the world’s most astute investors choose to pay attention.