AI‑directed bank accounts could move deposits quickly among banks, weakening a funding advantage that supports long‑term credit, according to a Federal Reserve Bank of Dallas analysis released August 25. Deposits typically remain at banks for years and usually earn lower rates than market rates, giving them a long effective duration that helps banks hold assets sensitive to interest‑rate changes. The analysis measured duration using weighted average life multiplied by one minus the deposit beta, which reflects how responsive deposit rates are to short‑term rate movements. Instant settlement and programmable AI rules could allow yield‑sensitive customers to shift balances rapidly. In June 2026, The Clearing House announced a plan to develop 24/7, interoperable tokenized commercial‑bank money for automated commerce. Using balance‑sheet data as of July 15 and its own duration assumptions, the Dallas Fed estimated about $7 trillion of asset‑side interest‑rate exposure in 10‑year equivalents, of which roughly $5.84 trillion derived from the duration characteristics of non‑time‑deposit balances. Under a scenario assuming a 10 percent increase in deposit price sensitivity and a four‑year weighted average life, the model showed a reduction of about $700 billion in duration‑risk appetite. A separate scenario with a 10 percent reduction in weighted average life cut projected maturity‑transformation capacity by about $580 billion. The model linked the $5.84 trillion of deposit‑backed assets to a $700 billion potential decline in duration‑risk appetite if deposit sensitivity rises. The analysis noted that banks might respond by issuing more term debt or increasing reserves and Treasury holdings, which could raise borrowing costs for consumers and businesses. A 2025 Central Bank of Brazil paper found that broader use of the Pix instant‑payment system increased liquid‑asset holdings and reduced liquidity transformation, illustrating how instant payment systems can affect bank liquidity behavior, though Pix is not a direct analogue to tokenized deposits. The authors emphasized that tokenized deposits are still early in development, the magnitude of the effect is uncertain, and the views expressed should not be attributed to the Dallas Fed or the Federal Reserve System.