US Banks: Agentic AI is a long-term bank-funding risk, but deposit betas—not app downloads—will determine whether disruption is real

Overview

This US banks commentary examines whether rapid adoption of Meta Platforms' Muse agent signals a threat to low-cost bank deposits. BofA’s central conclusion is that app traction establishes the opportunity for AI-driven cash optimization, while deposit betas will establish whether it is materially disrupting bank funding.

Core views

Bank stocks sold off as Muse’s rapid adoption renewed concerns that agentic AI could weaken the customer inertia supporting low-cost deposits. BofA argues that the relevant change is not faster money movement—consumers can already transfer funds—but lower decision latency. An agent could identify excess liquidity, compare yields and move cash without requiring the customer to initiate a separate decision. PayPal’s planned Muse checkout across its merchant network, Stripe Link integration and Shopify’s Shop Pay enablement demonstrate that major payment infrastructure can connect credentials and transactions to a third-party agent. The report views this as a precedent: the gap between an agent finding and purchasing a better product and finding a better yield and moving excess cash is narrowing. Early Muse use supports the existence of the consumer use case but does not yet prove financial disruption. Sensor Tower data cited by BofA show approximately 2.5 million cumulative US downloads and approximately 557,000 US daily active users as of September 19. On US iOS, Muse reached 1.43 million cumulative downloads by day 12, compared with 1.37 million for ChatGPT in a comparable post-launch period, and had a 4.9-star rating across more than 25,000 reviews. Users appear to value its ability to navigate websites, prioritize emails and analyze subscriptions with limited intervention. However, the report emphasizes that willingness to use an agent for general workflows is not the same as willingness to grant it broad access to financial accounts or delegate financial decisions. BofA expects the transition to be evolutionary rather than revolutionary. Privacy, trust, credential protection, authentication, consent, error resolution and liability for unauthorized transactions remain meaningful constraints, particularly given concerns about unpredictable AI behavior and the higher stakes of financial loss. Large corporate and high-net-worth balances are already managed relatively efficiently, so the report sees the greater exposure in excess consumer and small-business funds, especially non-interest-bearing checking balances kept for convenience, payments functionality, overdraft protection and access to bank services. Transactional deposits should remain necessary, but AI could gradually lower the structural level of idle balances, increase competition for marginal funding and reduce the value of customer inertia. The report regards frictionless sorting of excess liquidity into higher-yielding alternatives as a genuine threat to industry net interest margins through stickier or rising deposit costs. Still, it says the risk has a long tail: younger customers are more likely to adopt AI optimization, while the Great Wealth Transfer is expected to unfold over the next two decades. The report estimates that roughly 35% of deposits are stable consumer balances based on liquidity coverage ratio disclosures from the six largest US banks. It also notes that, despite the SVB failure and a 5.50% federal funds rate in 2023, industrywide deposit costs peaked at 2.52% in 3Q24, indicating that deposit economics reflect not only inertia but also customers’ implicit valuation of banking services. BofA’s operational test is deposit beta behavior, rather than downloads, daily active users or app-store rankings. The disruption case would gain evidence if deposit costs rise faster than can be explained by interest rates, normal competitive repricing or changes in deposit mix, alongside lower non-interest-bearing deposits and lower consumer checking balances per account. Until then, the report considers the threat largely conceptual, although it may still affect bank valuations. Strategically, BofA argues that banks should accept some self-cannibalization rather than let an external agent control both the recommendation and destination for customer funds. JPMorgan’s proposed Smart Cash product, intended to automate movement between checking and higher-yielding brokerage balances while managing cash flow, is presented as an example of internalizing optimization. Banks with integrated consumer banking, brokerage and wealth platforms may better retain assets within their ecosystems; less integrated banks may need to improve their own offerings, partner with third parties or accept greater funding and margin pressure. Over time, the report says banks may also need revenue models beyond spread income, including subscription or service fees, if customer interfaces shift to outside agents.

Analysis framework

The report starts with Muse’s adoption and payment-platform integrations, then distinguishes consumer interest from measurable bank-funding disruption. It assesses which deposit pools are most exposed, traces potential cash sorting through deposit costs and net interest margins, and identifies deposit betas and deposit mix as the required evidence. It concludes with strategic implications for banks’ control of the customer interface and balance ecosystem.

Methodology notes

• Other

  Deposit-beta and deposit-mix monitoring

  The report treats deposit betas—how quickly deposit costs rise relative to rate changes—together with non-interest-bearing deposits and checking balances as the practical test of whether AI is changing funding behavior.

• Financial-sector metricsNet Interest Margin (NIM) Analysis

  Deposit sorting and funding-cost transmission

  The report links more efficient movement of excess cash into higher-yielding alternatives to higher deposit costs and, all else equal, pressure on bank net interest margins.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• JPMorgan Chase

  Example of a bank pursuing internal balance optimization through its proposed Smart Cash product.

  Strengths

  An integrated consumer banking, brokerage and wealth ecosystem may help retain customer assets within the franchise.

  Comparison

  The report suggests banks without similarly integrated platforms may face greater funding and margin pressure.

  Risks

  External AI agents could control the recommendation and destination of customer funds.

Key data

• Muse cumulative US downloads~2.5 millionAcross platforms as of September 19
• Muse US daily active users~557,000As of September 19
• Muse US iOS cumulative downloads1.43 millionBy day 12, versus 1.37 million for ChatGPT over a comparable post-launch period
• Muse app rating4.9 starsAcross more than 25,000 reviews
• Estimated stable consumer deposits~35%Based on liquidity coverage ratio disclosures from the six largest US banks
• Industrywide deposit-cost peak2.52%Reached in 3Q24 after the federal funds rate reached 5.50% in 2023

Impact & implications

The report says AI agents could gradually make consumer and small-business excess cash more rate-sensitive, raising marginal funding competition and pressuring bank margins. It argues that the more immediate strategic issue is whether banks retain ownership of the agent, wallet and customer relationship, even if the financial impact takes years to emerge.

Risks

• Agentic AI could reduce excess balances in low-rate accounts, increase deposit betas and pressure bank net interest margins.
• Banks may lose product differentiation, cross-selling opportunities and customer-interface control if third-party agents become the primary financial decision channel.
• Broader adoption remains uncertain because users may resist granting third-party agents access to financial accounts amid privacy, security and liability concerns.

What to watch

• Non-interest-bearing deposits as a percentage of total deposits.
• Balance per consumer checking account.
• Deposit betas relative to prior rate-hike and rate-cut cycles.
• Whether deposit costs rise faster than can be explained by rates, ordinary competition or deposit-mix changes.