Your AI Saves You Money. Does It Drain Your Country’s Banks?

What Matters

The Bank for International Settlements published new evidence in October 2026 showing that stablecoins and bank deposits behave as partial substitutes. Higher bank deposit rates are followed by slower stablecoin growth, with the relationship stronger for USDC than USDT.

The IMF has separately warned that foreign-currency stablecoins could weaken domestic deposit bases, raise bank funding costs and reduce domestic credit, particularly in emerging markets where demand for dollar assets is already high.

The agent changes the speed of the problem

Traditional deposit flight has friction. Humans compare rates imperfectly, delay decisions, tolerate inconvenience and often leave money where it already sits.

An AI agent does not need to be loyal to a bank. It can be instructed to optimise for cost, yield, settlement speed, FX spread, liquidity or counterparty risk.

Stablecoins are not just another payment app

Stablecoins increasingly combine three functions that used to sit in separate systems: payments, dollar access and programmable settlement.

The BIS estimates that roughly 98% of stablecoin value is dollar-denominated. That matters especially outside the United States. A consumer in a weaker-currency economy is not merely choosing between two payment methods. They may be choosing between holding a domestic bank deposit and holding a digital claim linked to the U.S. dollar.

That turns payment optimisation into a currency-allocation decision.

Why local banks care

New York Fed evidence goes beyond theory

A 2026 Federal Reserve Bank of New York staff report found evidence that stablecoin activity can transmit liquidity shocks into the banking system. Banks associated with stablecoin deposits needed substantially larger reserve balances, while partner banks saw their loan share of assets contract relative to peers.

That does not prove stablecoins automatically destroy lending. It shows that the structure of the reserves matters. Money can leave one type of bank funding and reappear in a different, more liquidity-intensive form.

The emerging-market problem is sharper

The IMF’s 2026 work on foreign-currency stablecoins highlights a particular vulnerability in economies with unstable currencies, capital controls or restricted access to foreign assets.

Stablecoins can make dollar access easier. That can be useful for households and businesses protecting themselves from inflation or poor payment infrastructure. But it can also accelerate currency substitution and reduce the deposit base supporting local credit.

What AI adds: automatic dollarisation

The most important future scenario is not “people buy stablecoins.” It is “agents choose them automatically.”

Suppose an agent is instructed to keep business cash in the safest liquid asset with the lowest transfer cost and best purchasing-power protection. In a country with a weak currency and expensive banking rails, the agent may repeatedly prefer dollar stablecoins.

If that rule becomes embedded in payroll, treasury, freelancer tools, e-commerce, merchant agents and remittance systems, dollarisation can become software behaviour rather than an active human choice.

DN Agentic Deposit Drain Index

DN proposes a framework for estimating how exposed a banking system is to agent-accelerated deposit substitution.

DN Agentic Deposit Drain Calculator

This is not automatically bad for consumers

Stablecoin competition can improve financial services. Banks may respond by raising deposit rates, reducing payment fees, improving cross-border transfers or offering tokenised deposits.

The BIS working paper itself does not claim that stablecoin growth mechanically reduces aggregate bank deposits or lending. The effect depends on where reserves are held, how banks respond and whether stablecoins complement or replace existing money.

The competitive pressure can therefore be healthy at the household level even while creating difficult transition risks at the system level.

The banks’ counterattack: make deposits programmable

The likely response is not for banks to defend static accounts forever. It is to make bank money more competitive with stablecoins.

Central banks are already testing the agent problem

Project Agenti, launched by the BIS Innovation Hub with the Bank of Canada and Swiss National Bank, is explicitly modelling how AI-driven behaviour may affect payment timing, liquidity use, gridlock and shock propagation in wholesale payment systems.

That is an early sign of where policy is heading: central banks are not only asking how digital money changes payments. They are asking what happens when the decision-maker itself becomes software.

The risk is coordination without coordination

Financial crises often involve correlated behaviour. Everyone runs to cash, everyone sells the same collateral, everyone seeks the same safe asset.

AI agents introduce a new version of this problem. They do not need to communicate with one another to behave similarly. If they are trained on similar data, use similar optimisation objectives and respond to the same price signals, they may converge on the same action independently.

That can turn an apparently rational treasury rule into a system-wide liquidity event.

DN Alpha Thesis

Agentic finance could make money more efficient for individuals while making funding less sticky for institutions.

That trade-off is easy to underestimate because banking systems were built around slow human behaviour. Depositors are inattentive. Businesses tolerate poor treasury management. Cross-border payments are inconvenient. Switching is cumbersome.

Agents remove those frictions.

The banking question of the next decade may therefore be simple: what happens when deposits become as easy for software to move as packets are on the internet?

Sources

• BIS Working Paper 1382 — Are stablecoins and bank deposits substitutes? (Oct. 2026)
• BIS Paper 170 — Stablecoins and the international monetary system
• BIS Annual Economic Report 2026
• Federal Reserve Bank of New York — Stablecoin Disintermediation (2026)
• IMF — Foreign Currency Stablecoins: Macro-Financial Stability Considerations (2026)
• IMF — How Agentic AI Will Reshape Payments (2026)
• BIS — Payments in the Age of AI Agents
• BIS Project Agenti

FAQ

Can stablecoins reduce bank deposits?

They can act as partial substitutes for deposits, but the net effect depends on where reserves are held, bank pricing responses and whether users treat stablecoins as payment tools, savings assets or both.

Why would AI agents accelerate deposit outflows?

Agents can continuously compare yield, FX risk, fees and settlement speed, reducing the inertia that keeps human deposits sticky.

Are emerging markets more exposed?

Potentially. Countries with unstable currencies, capital controls or strong demand for dollar assets may face greater currency-substitution and bank-funding risks if foreign-currency stablecoins become easy to access.