Consumer AI agents and their implications for financial services: Consumer AI agents could disrupt financial-services friction, but payments may emerge as a beneficiary

Overview

This thematic note examines whether consumer AI agents could erode financial-services businesses that benefit from customer inertia and transaction friction. Bernstein concludes that the disruption case is real but incomplete: provider control over data and access, limited consumer trust, unresolved liability and regulation should constrain adoption, while payments infrastructure may be positioned to benefit.

Core views

Bernstein frames Meta's Muse as a pivotal consumer-AI moment. The agent rapidly reached No. 1 on the U.S. app store and recorded 2.8 million downloads, with users applying it to tasks ranging from appointments and reservations to insurance comparisons, subscription cancellations and customer service. Its emergence prompted a sell-off in financial-services businesses exposed to consumer inertia or friction: insurance, brokers, mortgage lenders and banks were down 5-15% since launch, while payments were comparatively spared. The report argues that an agent able to compare, switch, negotiate and move money could undermine card top-of-wallet habits, sticky deposits, broker cash-sweep balances, insurance-renewal pricing and some paid financial advice. However, Bernstein argues that the central question is not whether agents can execute financial tasks, but whether they can be trusted and permitted to do so at scale. Third-party agents require merchant and provider participation, while providers may resist if participation cedes checkout economics, the direct customer relationship or control of logins, identity checks, firm quotes and eligibility. Amazon's decision to block Muse and Insurify's concerns about agents presenting insurance quotes without critical coverage terms illustrate the likely response: more blocking, paid-access agreements, provider-owned agents and potentially curated access to what agents can see. Data access is a major control point for financial institutions. Bernstein notes that the CFPB's prior Open Banking 1033 rule, which would have required free secure API access to customer data for authorized third parties, was blocked. JPMorgan subsequently began charging data aggregators, citing API traffic that more than doubled over two years and elevated fraud associated with aggregator-initiated payments. A new Section 1033 proposal sent to White House review in August 2026 could permit banks to provide a specified number of free requests and charge thereafter, while potentially narrowing the definition of a consumer representative. The report notes that it has not seen evidence that the proposal specifically addresses AI agents. Consumer trust remains a meaningful adoption constraint. TD Bank's survey of more than 2,500 Americans found that 55% used AI to help manage finances, up from 10% a year earlier, and 62% considered AI capable of providing reliable information. Yet only 18% were comfortable allowing it to make important financial decisions independently. Other surveys similarly indicate a preference for recommendation over autonomy: only 7% of U.S. and UK shoppers in an ACI Worldwide/YouGov survey would allow an AI assistant to buy without approval; Accenture found 32% would permit an agent to decide within preset limits, but just 12% would let it decide independently at payment. Bernstein therefore sees trust, data access and regulation—not conventional checkout friction—as the principal obstacles to agentic adoption in financial services. The legal and regulatory framework is also lagging the technology. The report identifies unresolved questions over whether agent-submitted applications or purchases constitute valid consent, how providers can verify an agent's mandate, when optimization becomes regulated financial advice, whether agents require state licenses, whether data access can be charged for, and who bears liability if an agent exceeds a consumer's intent, uses stale data or recommends an unsuitable product. Incumbents may respond through lobbying, proprietary agents, favorable integrations or some degree of self-cannibalization to retain customers. Bernstein is notably constructive on payment networks. It argues that agentic commerce increases, rather than reduces, the value of Visa and Mastercard because consumers and merchants will need trusted credentials, dispute management, standards and protections. The networks collectively provide more than 7 billion credentials and access to more than 100 million merchants. The report expects agentic transactions to be tokenized and highlights Visa Intelligent Commerce and Mastercard Agent Pay, whose tokens add transaction context and assurance signals. In this view, greater digitization, transaction complexity and value-added services can shift the networks from an AI funding-source concern to AI beneficiaries. For payment service providers, the report sees merchant integration complexity as an opportunity for differentiation. Adyen and private Stripe can help merchants connect to multiple evolving agentic protocols; Adyen has launched Adyen Agentic and acquired Orb for AI usage-based billing. PayPal faces uncertainty over whether agents diminish the relevance of guest checkout, but Bernstein believes its two-sided wallet could expand its role in trust, protections, fraud and disputes, as well as savings, rewards and payment optimization. The report also suggests a deeper Google-PayPal integration could emerge as Google improves Spark within Gemini. BNPL providers Affirm and Klarna may be well positioned where transparent terms, 0% APR offers and distinctions between prime and near-prime or sub-prime borrowing matter. Overall, Bernstein expects the ecosystem to take time to settle and believes initial successful use cases may focus on solving high-friction journeys rather than replacing routine retail checkout or traditional financial services.

Analysis framework

Bernstein starts with the market reaction to Muse, then tests the disruption thesis through the mechanisms by which agents could reduce consumer friction. It weighs those effects against provider control of data and access, survey evidence on consumer trust, regulatory and liability constraints, and the different competitive roles of networks, processors, wallets and BNPL providers.

Methodology notes

• Event-Driven and Behavioral FinanceEvent-driven analysis

  Event-driven analysis of Muse's launch and the financial-services share-price reaction

  The report uses the launch of Muse as the triggering event, relating subsequent sector declines and changing market narratives to the expected effects of consumer agents.

• Competition & strategyValue chain analysis

  Financial-services value-chain and control-point analysis

  The report assesses who controls consumer authentication, data access, checkout, identity, quotes and transaction trust, and uses those control points to identify potential disruption and beneficiary roles.

Asset mapping & comparison

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

• Visa

  Potential beneficiary of agentic commerce through trusted, tokenized card payments and value-added services.

  Strengths

  Dispute management, standards, a large credential base and merchant acceptance.

  Comparison

  Bernstein groups Visa with Mastercard as networks of trust rather than merely settlement infrastructure.

  Risks

  The report notes market concerns that agentic commerce could disintermediate card networks.

• Mastercard

  Potential beneficiary of agentic commerce through trusted, tokenized card payments and value-added services.

  Strengths

  Dispute management, standards, a large credential base and merchant acceptance.

  Comparison

  Bernstein groups Mastercard with Visa as a likely agentic-commerce winner.

  Risks

  The report notes market concerns that agentic commerce could disintermediate card networks.

• Adyen

  Modern payment-service provider that can help merchants integrate evolving agentic-commerce protocols.

  Strengths

  Adyen Agentic and the acquisition of Orb for AI usage-based billing.

  Comparison

  Bernstein cites Adyen alongside private Stripe as a modern PSP able to navigate protocol complexity.

  Risks

  Agentic commerce remains in early, iterative phases with multiple competing protocols.

• PayPal

  Digital wallet that could expand its role in trust, protections, disputes and payment optimization.

  Strengths

  Two-sided wallet model, consumer protections, fraud handling, rewards and personalization capabilities.

  Weaknesses

  Agent-handled payments may reduce the relevance of guest-checkout friction.

  Comparison

  The report distinguishes PayPal's wallet opportunity from payment-network and processor roles.

  Risks

  The outcome depends substantially on strategic partnerships and agentic-commerce positioning.

• Affirm

  Potentially well positioned in agent-enabled BNPL comparison and financing selection.

  Strengths

  Relevant product transparency across near-prime and sub-prime interest-bearing loans and prime 0% APR loans.

  Comparison

  Discussed alongside Klarna as a potential BNPL beneficiary.

• Klarna

  Potentially well positioned in agent-enabled BNPL comparison and financing selection.

  Strengths

  Relevant product transparency across near-prime and sub-prime interest-bearing loans and prime 0% APR loans.

  Comparison

  Discussed alongside Affirm as a potential BNPL beneficiary.

Key data

• Muse downloads2.8MDownloads shortly after launch; Muse reached No. 1 on the U.S. app store.
• Financial-services share-price reaction5-15% declineInsurance, brokers, mortgage lenders and banks were down since Muse launched, according to the report.
• AI use for financial management55%TD Bank's 2026 survey of more than 2,500 Americans, up from 10% a year earlier.
• Comfort with autonomous financial decisions18%Share of TD Bank survey respondents comfortable allowing AI to make important financial decisions independently.
• Autonomous purchase approval7%Share of U.S. and UK shoppers in the June 2026 ACI Worldwide/YouGov survey who would let an AI assistant buy without approval.
• Visa and Mastercard credentials7B+Credentials cited by Bernstein as part of the networks' trust and acceptance infrastructure.
• Visa and Mastercard merchant acceptance100m+ merchantsMerchant footprint cited by Bernstein in its case for cards in agentic payments.

Impact & implications

Bernstein argues that agentic AI could pressure businesses that rely on friction, inertia or opaque pricing, but it is unlikely to create immediate broad-based disintermediation in financial services. Providers retain meaningful access and identity controls, while trust and regulatory requirements favor established payment infrastructure; the report sees payment networks and selected processors as better positioned than the market's initial doomsday framing implies.

Risks

• Consumer agents could erode card top-of-wallet habits, sticky deposits, broker sweep balances, insurance-renewal pricing and some paid financial-advice revenue.
• Unresolved consent, authorization, identity verification, licensing, advice and liability rules could complicate agentic financial transactions.
• Providers may block agents, limit accessible information or charge for data access, slowing ecosystem development.

What to watch

• Whether merchants, banks, brokers, insurers and marketplaces block agents, enter paid-access arrangements or launch proprietary agents.
• The final CFPB Section 1033 proposal, including data-access fees and the definition of an authorized consumer representative.
• Evidence that consumers move from AI-assisted recommendations to autonomous financial decisions or payments.
• Adoption of agentic payment tokens and services such as Visa Intelligent Commerce, Mastercard Agent Pay and Adyen Agentic.
• Potential strategic integrations between consumer AI platforms and payment providers, including a deeper Google-PayPal relationship.