AI Agents Could Drain Cheap Bank Deposits and Reshape Financial Rails

What to Know

• Apollo Chief Economist Torsten Slok warned that agentic AI assistants could trigger a slow-motion bank run by optimizing where households keep their cash.
• The concern centers on low-interest checking accounts, where the national average is cited at 0.1%, compared with higher-yield alternatives paying between 3.3% and 5% annual interest on deposits.
• Slok argued that banks rely on cheap deposits to fund loans, meaning a broad shift of household cash could create stress across the financial system.
• Examples cited in the market discussion include Revolut, SoFi, Varo, LendingClub and Wealthfront as firms paying higher deposit rates.
• A $10,000 balance could generate about $100 in annual interest in a higher-yield account, compared with about $10 from a checking account.
• Agentic finance refers to AI systems that do more than answer questions, including monitoring balances, comparing yields, moving idle cash and returning funds in time for bills.
• Mordor Intelligence estimates agentic AI in financial services at $7.78 billion in 2026 and projects $43.52 billion by 2031, while MarketsandMarkets sizes the narrower AI agents segment at about $845 million in 2025.
• Crypto payment infrastructure is already supporting agentic finance, with Coinbase’s x402 protocol enabling AI agents to pay for online services in stablecoins within seconds.
• The x402 Protocol has reportedly processed approximately 188 million to more than 205 million cumulative transactions, with around 69,000 active agents.
• Cloudflare, Google, Visa, Mastercard, AWS, Circle and Stripe have joined the x402 Foundation, now governed by the Linux Foundation.

AI Cash Managers Move From Concept to Systemic Question

Artificial intelligence is moving deeper into personal finance, and market participants are beginning to ask whether automated money assistants could change the deposit base of the banking system. Apollo Chief Economist Torsten Slok has warned that agentic AI tools may create a slow-motion bank run by shifting household cash from low-interest checking accounts into higher-yield alternatives without consumers needing to manually compare rates or initiate transfers.

The concern is not that depositors suddenly line up outside branches. The risk is more subtle. If AI assistants are granted permission to manage household cash, they could continuously scan balances, evaluate available yields and move idle funds to the institution offering a better return. That behavior, repeated across many households, could reduce the pool of inexpensive deposits that banks have traditionally used to support lending activity.

Slok’s warning arrives as agentic AI becomes a more prominent theme in finance and technology. Unlike chatbots that simply respond to questions, agentic systems are designed to take action. In household finance, that could mean paying bills, monitoring due dates, finding higher deposit rates, transferring cash between accounts and ensuring enough money returns to checking accounts before scheduled obligations come due.

The Deposit Rate Gap Is the Core Incentive

The basic driver behind the warning is the gap between what many checking accounts pay and what some higher-yield deposit products offer. Slok cited the 0.1% national average on checking accounts, while pointing to firms such as Revolut, SoFi, Varo, LendingClub and Wealthfront paying between 3.3% and 5% annual interest on deposits. For households, that spread creates a clear incentive to move cash away from accounts that pay very little.

The difference becomes more visible when applied to a simple cash balance. A $10,000 balance could generate about $100 in annual interest in a higher-yield account, compared with about $10 from a checking account. The dollar amounts may appear modest at the individual level, but the systemic concern is what happens if AI makes the optimization automatic, widespread and continuous.

Many consumers leave money in checking accounts because comparison shopping is time-consuming, transfers can be inconvenient and financial inertia is powerful. Agentic AI directly challenges that inertia. A personal finance assistant that can identify a higher return and execute a transfer may reduce friction enough to make deposit movement a routine background process.

Why Banks Depend on Cheap Deposits

Banks use deposits as a key source of funding for loans. Low-cost deposits are especially valuable because they allow lenders to extend credit while maintaining a spread between what they pay depositors and what they earn from borrowers. If cheap deposits leave the system or migrate to institutions paying much higher rates, banks may face pressure on funding costs and lending capacity.

Slok’s argument is that if every household used AI agents to optimize returns on cash balances, banks could lose a large share of the cheap deposits they rely on to make loans. That would not necessarily look like a classic panic, but it could still affect the broader financial system by changing the economics of credit creation.

The phrase slow-motion bank run captures this distinction. A traditional bank run is often associated with fear and urgency. An agentic version could be driven by optimization rather than panic. Funds would move not because depositors believe a bank is unsafe, but because automated tools determine that better yields are available elsewhere.

Agentic Finance Expands the Role of Automation

Agentic finance describes AI that acts rather than simply advises. These systems can monitor balances in real time, compare returns across institutions, move idle cash into higher-yield accounts and move it back in time for bills. In theory, the user sets broad preferences, while the agent handles routine execution.

That model could appeal to households seeking convenience. It could also appeal to financial platforms competing for deposits, payments volume and customer engagement. The more capable these agents become, the more they could reshape where money sits and how quickly it moves through the financial system.

Estimates of the market vary widely, reflecting how early the sector remains. Mordor Intelligence puts agentic AI in financial services at $7.78 billion in 2026 and projects $43.52 billion by 2031. MarketsandMarkets sizes the narrower AI agents segment at about $845 million in 2025. Those figures point to a market that is still developing, but one that many technology and finance firms are positioning around.

Crypto Rails Enter the Agentic Finance Debate

Crypto infrastructure is already playing a role in the discussion because agentic systems need fast, programmable payment rails. Coinbase’s x402 protocol has emerged as a widely used agentic payment standard, allowing an AI agent to pay for online services in stablecoins within seconds, without an account, card or human approval.

For AI agents, the ability to execute small payments quickly and programmatically is important. Traditional payment systems were designed primarily around human users, cards, accounts and established identity workflows. Agentic commerce raises a different set of requirements, including machine-to-machine payments, automated authorization and settlement that can occur as part of a digital workflow.

The x402 Protocol has reportedly processed approximately 188 million to more than 205 million cumulative transactions, with around 69,000 active agents. Those figures suggest that agentic payment activity is no longer only a theoretical concept. It is already being tested and used across digital environments where AI agents can interact with services and settle payments in stablecoins.

Major Technology and Payments Firms Join the Infrastructure Race

The institutional roster around the x402 Foundation underscores how seriously large firms are taking the agentic payments theme. Cloudflare, Google, Visa, Mastercard, AWS, Circle and Stripe have joined the foundation, which is now governed by the Linux Foundation. That combination of cloud, payments, stablecoin and internet infrastructure participants signals broad interest in standards for machine-driven financial activity.

For traditional finance, this development is both an opportunity and a challenge. Banks, card networks and fintech firms may benefit from new transaction flows if they adapt to agentic use cases. At the same time, the rise of programmable payment rails could increase competition for deposits, payments and customer relationships.

Nate Geraci, co-founder of the ETF Institute, has previously said that AI and crypto are both coming for the traditional banking model. The point reflects a broader market view that artificial intelligence may change how financial decisions are made, while crypto infrastructure may change how digital value is transferred. Together, the two trends could put pressure on legacy banking assumptions.

What It Could Mean for Households and Lenders

For households, agentic finance could make cash management more efficient. Instead of accepting low yields through inattention or convenience, consumers may be able to delegate routine optimization to software. The benefit would be higher potential interest income and less manual effort, provided users understand the risks, permissions and terms attached to each product.

For banks, the issue is more complicated. If depositors become more rate-sensitive because AI makes switching easier, lenders may need to compete more aggressively for balances. That could compress margins, alter loan pricing or change the structure of retail banking relationships. The risk highlighted by Slok is not just deposit movement, but the possibility that a foundational funding source becomes less sticky.

Regulators and financial institutions are likely to watch the trend closely as agentic tools move from early adoption toward broader household use. Questions around authorization, disclosures, liquidity, operational resilience and consumer protection may become more important if AI agents begin executing financial decisions at scale.

A Slow Shift With High Stakes

The debate over agentic bank runs remains forward-looking, and the outcome is not predetermined. Many households may be slow to grant AI tools broad authority over their money. Banks may respond with better rates, improved account features or their own agentic services. Fintech firms may also face trust and regulatory hurdles as they seek to manage more consumer cash.

Still, the warning from Slok highlights a critical point for the financial industry: automation can change behavior by removing friction. If AI assistants make it easy for households to pursue better returns on idle cash, deposits that once seemed stable may become more mobile. In that environment, the battle for consumer balances could become faster, more competitive and increasingly shaped by software.

Frequently Asked Questions (FAQs)

What did Torsten Slok warn about?

Torsten Slok warned that agentic AI assistants could trigger a slow-motion bank run by automatically moving household cash from low-interest checking accounts into higher-yield alternatives.

Why could AI agents pressure banks?

AI agents could pressure banks by reducing the amount of cheap deposits available to fund loans. If many households move cash to higher-yield products, banks may face higher funding costs and broader balance sheet pressure.

What is agentic finance?

Agentic finance refers to AI systems that can take action on behalf of users. In personal finance, that may include monitoring balances, comparing yields, moving idle cash and returning funds in time for bills.

What checking account rate was cited?

The national average on checking accounts was cited at 0.1%, creating a large gap with higher-yield deposit products that may offer more attractive returns.

Which firms were mentioned as offering higher deposit rates?

Firms mentioned in the market discussion include Revolut, SoFi, Varo, LendingClub and Wealthfront, with deposit rates cited between 3.3% and 5% annual interest.

How much interest could a $10,000 balance earn?

A $10,000 balance could generate about $100 in annual interest in a higher-yield account, compared with about $10 from a checking account, based on the cited comparison.

How is crypto connected to agentic finance?

Crypto is connected through payment infrastructure that can support machine-driven transactions. Coinbase’s x402 protocol allows AI agents to pay for online services in stablecoins within seconds.

What activity has the x402 Protocol reportedly processed?

The x402 Protocol has reportedly processed approximately 188 million to more than 205 million cumulative transactions, with around 69,000 active agents.

Why are major firms joining the x402 Foundation?

Major technology, payments and crypto firms are joining because agentic payments may become an important part of digital commerce, requiring standards for AI-driven transactions and programmable settlement.