Agentic commerce: Is Payments the winner this time?

Bernstein argues that agentic commerce, currently less than 1% of e-commerce activity, is a net positive rather than a disintermediation threat for Visa, Mastercard and Adyen. The firm expects 100% of agentic transactions to be tokenized and says cards retain structural advantages through dispute management, established standards, more than 7 billion credentials and over 100 million merchants. GenAI referrals reached 39% of Wayfair referrals and 28% for Walmart as of August 2026, but still account for only 1-2% of these retailers' total traffic; Bernstein rates Visa, Mastercard and Adyen outperform.

Analysis

The investable effect is not incremental payment volume alone; it is control of authentication, token lifecycle management, routing and dispute resolution as agents create more non-human purchase attempts. V and MA should defend network economics if credential-on-file transactions migrate to network tokens, but the larger upside accrues over 6-18 months only if tokenization increases value-added-services revenue and reduces fraud/authorization losses rather than merely shifting existing card volume into a new checkout interface.

ADYEN is the cleaner idiosyncratic beneficiary because fragmented agent channels raise integration and orchestration costs for global merchants, favoring a single-stack acquirer with enterprise distribution. The near-term risk is that agent-driven shopping increases price transparency and purchase reversals, compressing merchant conversion economics before it generates material incremental GMV; this is particularly unfavorable to lower-differentiation retailers and marketplaces such as TGT, ETSY and EBAY. High AI-referral shares are not yet proof of monetizable demand, and could instead represent low-intent research traffic.

Consensus is prematurely framing this as a payments-network winner. The more consequential 1-3 month variable is consumer delegation trust and liability allocation: a high-profile erroneous purchase, fraud event, or chargeback dispute could slow adoption and force networks, issuers and merchants to absorb more operating cost. Conversely, disclosed enterprise merchant pilots, higher tokenized-commerce penetration, or evidence of better authorization rates would validate a durable processing and network-revenue uplift.

Key Decisions for Investors

• Maintain a modest long V / MA basket for 6-18 months, but do not chase a headline-driven move; add only on market weakness. Thesis requires tokenized credentials and value-added-services growth to outpace baseline payment-volume growth over the next 2-3 earnings cycles.
• Initiate a 3-6 month long ADYEN versus short GPN pair, sized small: Adyen has greater upside to enterprise channel complexity while legacy acquirers face higher integration burden. Target 15-20% relative upside; exit if Adyen fails to cite agent-channel merchant adoption or if take-rate pressure accelerates.
• Avoid treating WMT, TGT, ETSY and EBAY AI referral metrics as a demand catalyst. Set an alert for disclosed conversion rate, basket size and return/chargeback data; without those metrics, referral growth is more likely to signal comparison-shopping pressure than revenue upside.
• Monitor V/MA quarterly disclosures for tokenized transaction mix, authorization uplift and fraud-loss trends. A deterioration in chargebacks or issuer incentives to favor account-to-account rails would falsify the network-margin thesis and warrants reducing exposure.