It turns out UBS has been using artificial intelligence not to trade markets, but to listen to them.
In a note entitled Deep Speak, the bank’s strategists have built an AI model that analyses how the tone of the world’s most powerful central bankers has shifted... and the results suggest a quiet but notable turn in the global monetary mood.
Using large language models, the system sifts through two decades of speeches, interviews and press conferences from the Federal Reserve, European Central Bank and Bank of Japan.
Each comment is scored on a hawkish–dovish scale and categorised by topic, such as inflation, employment or interest rates. The dataset updates daily and is now accessible through UBS’s “Evidence Lab” platform.
The findings show the Fed has moved further into dovish territory since September’s policy meeting, as concerns over growth and the labour market have overtaken inflation worries.
Jay Powell’s tone has softened, while Governor Philip Jefferson has emerged as one of the more dovish voices. Inflation, UBS notes, “no longer offsets the dovish shift”, a signal that rate cuts could move back into focus if the economy continues to cool.
At the European Central Bank, by contrast, tone has steadied close to neutral. Christine Lagarde’s language is balanced, Isabel Schnabel’s mildly dovish, and Olli Rehn stands out as the lone hawkish outlier.
Inflation is now “only a marginal driver” of the ECB’s messaging, with more attention turning to interest rates themselves, a sign that officials are increasingly comfortable with where policy stands.
In Tokyo, the pendulum swings the other way. The Bank of Japan’s tone has turned distinctly firmer since the summer, with Governor Kazuo Ueda leading a cautious hawkish shift focused on the labour market and gradual balance sheet tightening.
UBS’s model suggests the global policy cycle is diverging again: a softer Fed, a steady ECB and a more assertive BoJ. For investors parsing every speech for hidden clues, the machines may already be one step ahead.