Agentic commerce is still in an iterative stage, and payment networks and leading platforms have clearer phased upside

Overview

This report summarizes the cross-industry discussions from Bernstein's inaugural Agentic Commerce Day in New York, with participants spanning Wayfair, Mastercard, BCG, former Walmart executives, DTC brands, Koio, Paxos, and others. The report argues that agentic commerce is still in the product-market-fit iteration phase. The most realistic near-term outcomes are AI-enhanced product discovery, conversational shopping, merchant catalog optimization, marketing efficiency, and customer service efficiency, rather than agents fully completing complex shopping journeys on behalf of consumers.

Core views

The core views are: first, AI shopping is changing the discovery entry point more than it is immediately replacing consumer transaction behavior; second, Google has a good starting position because of its shopping graph and underlying product data, while OpenAI's abandonment of Instant Checkout highlights the difficulty of changing consumer behavior and connecting checkout infrastructure; third, merchants must maintain high-quality, real-time, AI-readable catalog, inventory, promotion, and rewards information; fourth, whether retail media budgets migrate depends on whether AI shopping experiences can improve conversion and create habits; fifth, Visa/Mastercard and similar payment networks may become key beneficiaries of the risk, liability, and tokenization standards required for agentic transactions.

Analysis framework

The report uses a thematic conference-note and cross-industry expert-interview format, placing e-commerce, retail media, payment networks, stablecoins, DTC brands, and large retail platforms within a single framework to assess the commercialization path, bottlenecks, and investment implications of AI agents moving from product discovery to transaction execution.

Methodology notes

• Business model analysisAgentic Commerce adoption path

  Staged evolution from discovery to transaction

  The report distinguishes between short-term feasible AI product discovery and conversational shopping, and the longer-term fully agentic ordering flow, emphasizing that the latter requires mature checkout, inventory, fulfillment, liability, and payment infrastructure.

• Channel and traffic analysisSEO to GEO

  Generative engine optimization

  The report argues that many of the factors driving SEO will also affect GEO, and merchants need to maintain structured, real-time, high-quality product catalogs and descriptions so products can be discovered by AI shopping entry points such as ChatGPT and Gemini.

• Payment infrastructure analysisTokenization and agentic payment standards

  Security, consent, identity, audit, and dispute management

  Agentic transactions will amplify liability and risk issues, so the report believes card networks have an advantage in setting standards, providing tokens, managing risk, and delivering value-added services.

Asset mapping & comparison

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

• Visa Inc / Mastercard Inc

  Potential primary beneficiaries

  Strengths

  They have payment network standards, tokenization, identity verification, dispute management, risk control, and value-added service capabilities; they also have room to participate in Agentic Commerce and stablecoin-linked cards.

  Weaknesses

  Realized benefits depend on actual scale adoption of agentic transactions and the rollout of ecosystem standards.

  Comparison

  Compared with a single merchant or wallet, network companies are more like joint standard-setters.

  Risks

  Liability allocation, regulatory requirements, the pace of consumer behavior change, and competition from alternative payment infrastructure.

• Large marketplaces and leading retailers

  Phased beneficiaries or defensive beneficiaries

  Strengths

  They have traffic, brand strength, fulfillment networks, pricing power, product catalogs, and resources for AI experimentation, and can integrate directly with AI labs.

  Weaknesses

  If AI changes the search entry point, platforms may face pressure on traffic sources and retail media budget reallocation.

  Comparison

  Compared with mid-sized merchants, large platforms have more ability to build proprietary AI shopping experiences.

  Risks

  Consumer behavior shifts, AI entry points taking commission or diverting traffic, and advertising budget leakage.

• DTC brands

  Short-term beneficiary

  Strengths

  AI tools may improve discoverability for niche products in specialized categories, and some DTC brands can receive as much as about 10% of referral traffic from AI chatbots.

  Weaknesses

  Fulfillment, pricing, brand trust, and data infrastructure are usually weaker than those of large platforms.

  Comparison

  Discovery efficiency may improve in the short term, but large brands and platforms may catch up over time.

  Risks

  Discovery traffic does not necessarily convert to transactions, and it can be copied or suppressed by large platforms.

• Wayfair Inc

  AI-assisted shopping case study and covered name

  Strengths

  The report notes that Wayfair is building an AI-assisted shopping experience and has launched UCP in partnership with Gemini.

  Weaknesses

  Higher-consideration categories have more complex shopping journeys, and fully agentic transactions are still a long way off.

  Comparison

  The disclosed rating is Market-Perform, which is more neutral than the positive ratings on payment networks.

  Risks

  AI experience investment may not convert as expected, traffic entry points may shift, and the home category has a long decision cycle.

• Stablecoins and stablecoin-linked cards

  Payment infrastructure extension opportunity

  Strengths

  Card networks provide acceptance, standards, and monetization models, and stablecoin-linked cards may be more practical in C2B scenarios.

  Weaknesses

  Machine payments and microtransaction business models are still at a very early stage, and infrastructure and orchestration layers are still being built.

  Comparison

  Compared with pure on-chain payments, card-network solutions are closer to the existing merchant acceptance stack.

  Risks

  Regulation, settlement, on/off ramps, user experience, and uncertainty around network companies' business models.

Key data

• Share of AI chatbot referral trafficBelow 1% for large platforms; up to about 10% for some participating DTC brandsThis shows that AI discovery has emerged but has not yet become a mainstream transaction entry point.
• Share of merchants advancing agentic commerce discovery work70% are already advancing, and BCG expects this to rise to 90%Merchant attention is relatively high, with a focus on product catalogs, AI visibility, and marketing efficiency.
• Tokenization view on agentic transactionsThe report says 100% of agentic transactions may need to be tokenizedThe reason is increased complexity around intent, disputes, audit trails, identity, and authentication.
• Wayfair ratingMarket-PerformThe report lists W as Market-Perform within US Emerging Internet coverage.
• Payment stock ratingsV, MA, Adyen, TOST, and XYZ are rated Outperform; FISV, FIS, GPN, PYPL, and KLAR are rated Market-PerformThis reflects the report's relatively positive view of payment networks and selected payment companies.

Impact & implications

In investment terms, Agentic Commerce should not be viewed in the short term as a force that will immediately disrupt the existing e-commerce transaction chain. The more investable themes are AI-driven product discovery, retail media conversion improvement, merchant catalog and data infrastructure, and payment tokenization plus value-added services. Visa/Mastercard have a strong beneficiary thesis through standards setting, tokens, risk control, dispute handling, and data value-added services; leading marketplaces that can convert AI shopping experiences into higher conversion and stronger user habits may retain and expand retail media budgets; DTC brands may gain better discoverability in the near term, but in the long run large platforms and strong-fulfillment brands may regain the edge.

Risks

• Consumers may continue to maintain their relationships with existing marketplaces and retailers, causing agentic transaction adoption to be slower than expected.
• Infrastructure issues such as checkout conversion, real-time pricing, after-sales support, inventory catalog synchronization, fulfillment, and multi-cart support may limit Agentic Commerce scale-up.
• If AI shopping entry points reallocate search traffic, retail media budgets and platform economics may become uncertain.
• Liability, dispute, identity verification, and audit issues arising from agentic recommendations or delegated purchasing have not yet formed clear industry standards.
• Mid-sized merchants may lag because of insufficient technology, talent, and integration capability.
• Stablecoins and machine payments are still early, and their business models and regulatory paths remain unproven.

What to watch

• Whether the share of referral traffic from AI shopping entry points such as ChatGPT and Gemini rises meaningfully from the low-single-digit range.
• Whether industry consensus emerges among OpenAI, Google, Stripe, Visa/Mastercard, and merchants on payment liability, authentication, and token standards.
• Progress in integrating merchant product catalogs, inventory, promotions, and rewards programs with AI shopping experiences.
• Whether AI-assisted shopping improves conversion and supports retail media budget retention or expansion.
• User adoption and commercial conversion of proprietary AI shopping experiences at large retailers such as Wayfair.
• The development of stablecoin-linked cards, B2B/B2C/P2P settlement, and on/off-ramp infrastructure.