Agentic commerce: Goldman Sachs sees agentic commerce as a gradual but broad reshaping of discovery, payments and commerce infrastructure.

Overview

This deep dive examines how action-oriented AI agents could change the consumer shopping journey, beginning with discovery and eventually extending to purchase and payment. Goldman Sachs expects a multi-year transition rather than an immediate disruption, with value increasingly accruing to businesses that control consumer intent or provide the trusted infrastructure needed to convert it into transactions.

Core views

Goldman Sachs frames agentic commerce as the next potential stage of eCommerce adoption: AI agents are already used for product discovery, research and recommendations, but autonomous purchasing remains early. The firm expects progress over the next 3–5+ years and potentially longer, comparing the path to the gradual development of eCommerce rather than an overnight shift. It estimates that roughly 60% of retail and services spend, or more than $12.3tn, falls into high- and medium-likelihood categories. Within this, $2.6tn of US consumer spending is identified as a near-term opportunity in high-likelihood categories; about 2% agentic penetration of high-likelihood card-present spending could add approximately 1 percentage point to eCommerce growth. Low-risk, recurring and commoditized purchases should adopt first, while subjective categories such as apparel and luxury should move more slowly because the cost of misreading consumer intent is higher. The report sees a shift from conversational AI toward action-oriented personal agents that aggregate consumer intent across discovery, comparison, checkout and payment. Technology remains fragmented: agents can work through browser-based flows, embedded merchant tools, stored-wallet systems or operating-system-style personal agents. Merchant catalog exposure, shopping protocols and payment authorization must work together. MCP is described as an integration layer for catalog and inventory access, while UCP supports commerce interactions across discovery, checkout and post-purchase workflows. Visa Trusted Agent Protocol and Mastercard Agent Pay are intended to add agent verification, consumer intent and payment authorization. Goldman Sachs argues that protocols offer a route to scale, but clearer stakeholder rules and liability allocation are still required. The report expects advertising monetization to migrate with commercial intent rather than disappear. AI already improves creative generation, targeting, campaign optimization and measurement; agentic commerce is a longer-duration change that could move discovery from search engines and retailer sites into AI interfaces. Sponsored recommendations, product-feed advertising and agent-facing formats could emerge, similar to how the desktop-to-mobile transition changed the value of user relationships. META and GOOGL are viewed as primary secular beneficiaries because they combine consumer distribution, AI capabilities, data, commerce integration and advertising ecosystems. META's Muse had more than 2.8 million downloads after launch and ranked as the top free iOS app in the US in its second week, according to Sensor Tower cited by the report. Still, attribution, measurement and third-party merchant participation remain unresolved. For retailers and marketplaces, the report identifies a tension between broader distribution and retaining control of the customer relationship. AI agents could surface smaller merchants and support Shopify's ecosystem, yet optimization around price, speed, availability and transaction reliability could favor scaled operators such as Amazon and Walmart. Amazon's blocking of unauthorized Muse shopping access illustrates retailer concerns over first-party data, loyalty, retail-media economics and margin control. Walmart's Sparky is cited as an example of a retailer-controlled agent; the report states weekly active customers doubled year over year, grew 60% quarter over quarter, and generated a 40% higher average order value. Shopify is viewed as a particularly well-positioned infrastructure beneficiary because catalog management, inventory synchronization, checkout, Shop Pay, fulfillment, identity and returns remain necessary regardless of where discovery occurs. Goldman Sachs sees card networks Visa and Mastercard as well positioned because agentic payments are likely to build on existing card infrastructure rather than new payment forms. Network effects, tokenization, authentication, fraud prevention and real-time decisioning could support transaction volume and value-added-services attachment. More than 50% of V/MA eCommerce transactions are tokenized, according to the report, and Visa estimated roughly $650 million of fraud reduction in FY24 from tokenization. The firm also argues that smaller baskets and fragmented purchases could raise network yield through fixed per-transaction fees. It estimates that a 5% acceleration in V/MA cybersecurity-related value-added services could accelerate VAS growth by 90bps for Visa and 200bps for Mastercard, and total net revenue growth by 20bps and 80bps, respectively. The threat environment may grow faster than legitimate autonomous commerce. Authorized agents can resemble malicious automation because cloud infrastructure, shared IP addresses, programmatic navigation and machine-speed activity weaken the usefulness of traditional fraud signals. The report expects demand for verified agent identity, consumer-agent linkage, merchant verification, fraud decisioning, cybersecurity and risk controls to rise. It estimates cybersecurity spending currently represents roughly 1–2% of eCommerce revenue based on a 75bps retailer cybersecurity-spend estimate and assumptions about the online share of that budget, and expects higher intensity over time. Cloudflare could benefit through bot management, API security and application security, although Goldman Sachs says the incremental revenue opportunity remains unclear. For EFX, TRU and FICO, agentic commerce is described as longer-duration optionality rather than a quantified earnings driver. EFX has broad exposure to consumer identity, merchant verification, income, employment and credit inputs; TRU is differentiated in persistent identity linkage and digital-network risk; and FICO has its clearest potential monetization path in fraud and credit-decisioning software if automated financing and payment decisions become more frequent or complex. The report stresses that payment networks, issuers, wallets and agent platforms will still control credentials, tokenization and authorization, so the opportunity is concentrated in higher-risk events such as unfamiliar merchants, new account linkages, financing requests and changes in delegated authority. In live-event ticketing, Goldman Sachs expects primary platforms with differentiated inventory to fare better than secondary marketplaces. Agents could improve event discovery, conversion and sell-through, particularly for unsold capacity; Live Nation has stated that 95% of concerts do not sell out, with 60–70% sell-through at amphitheaters and 65–75% at theaters. However, easier price comparisons could commoditize secondary inventory, pressure service fees, weaken ancillary-product attachment and ultimately replace search costs with new AI-platform referral or placement fees. The report therefore sees a more favorable setup for primary ticketing platforms than for secondary marketplaces.

Analysis framework

Goldman Sachs traces the shopping journey from discovery to checkout, payment and post-purchase activity, then assesses how the shift changes economics for advertising, retailers, payments, cybersecurity, identity, credit decisioning and ticketing. It uses historical eCommerce adoption, category-level transaction risk, infrastructure and protocol analysis, company capability comparisons, and selected scenario sensitivities to identify likely beneficiaries and constraints.

Methodology notes

• Industry AnalysisSupply-demand framework

  Category-based agentic-commerce adoption framework

  The report sizes addressable spending and classifies purchase categories by transaction risk, complexity and subjectivity to determine where agentic adoption may occur first.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Commerce-stack transmission analysis

  The report follows effects from consumer discovery through merchants, advertising, payment networks, identity providers, fraud controls and lenders.

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Amazon sum-of-the-parts valuation

  Amazon's stated 12-month target is based on separate valuation multiples for North America, International and AWS.

• Valuation methodsDCF (Discounted Cash Flow)

  Modified discounted cash flow valuation

  The report uses modified DCF approaches alongside trading multiples for several covered companies, including GOOGL, META and STUB.

Asset mapping & comparison

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

• Meta Platforms (META)

  Potential beneficiary through Muse, consumer distribution and AI-native commerce and advertising monetization.

  Strengths

  Large user base, established consumer relationships, advertising platform and third-party commerce integrations.

  Weaknesses

  End-to-end scaling depends on permissions, account connectivity and merchant participation.

  Comparison

  Viewed with GOOGL as a primary digital-advertising beneficiary.

  Risks

  Competition for engagement and advertising dollars, investment pressure and regulatory scrutiny.

• Alphabet (GOOGL)

  Potential beneficiary through search, Gemini, AI Mode, commerce protocols and aggregation of commercial intent.

  Strengths

  Extensive distribution, AI infrastructure, compute capacity and advertiser relationships.

  Weaknesses

  Discovery migration may require continued evolution of monetization formats.

  Comparison

  Viewed with META as a primary secular beneficiary in digital advertising.

  Risks

  Search disruption, competitive utility, advertising-dollar competition and regulation.

• Shopify (SHOP)

  Potential beneficiary as merchant infrastructure for fragmented AI-driven commerce.

  Strengths

  Catalog, inventory, checkout, Shop Pay, fulfillment and merchant workflow infrastructure; role in UCP and Agentic Storefronts.

  Comparison

  Contrasted with legacy commerce platforms that may face more urgency to modernize.

  Risks

  Consumer weakness, investment-driven margin dilution and competition.

• Visa (V) and Mastercard (MA)

  Potential beneficiaries from agentic payment volume, tokenization, fraud controls and value-added services.

  Strengths

  Card-network scale, tokenization, authentication, fraud prevention and network effects.

  Weaknesses

  Liability and authorization frameworks remain unfinished.

  Comparison

  Viewed as better positioned than a new payment-form-factor model because agentic payments are expected to build on existing card rails.

  Risks

  Macro weakness, cross-border recovery, competition and regulation.

• Equifax (EFX), TransUnion (TRU) and Fair Isaac (FICO)

  Potential providers of identity, merchant-risk, fraud and credit-decisioning inputs for higher-risk agentic events.

  Strengths

  EFX has broad verification data; TRU has persistent identity and network-risk capabilities; FICO has decisioning and fraud software.

  Weaknesses

  Potential revenue is optionality and depends on incremental paid decisions rather than migration of existing workflows.

  Comparison

  FICO is viewed as having greater potential in software decisioning than incremental Scores volume.

  Risks

  Adoption may not create incremental checks, and AI competition or macro and credit conditions may weigh on demand.

• Cloudflare (NET)

  Potential beneficiary from agent verification, bot management, API security and application security.

  Strengths

  Sits in front of merchant websites and APIs; can distinguish authorized agents from malicious bots.

  Weaknesses

  Incremental revenue opportunity remains unclear.

  Comparison

  Part of the report's broader fraud and cybersecurity beneficiary set.

• Walmart (WMT) and Amazon (AMZN)

  Potential retail beneficiaries where agents optimize for price, speed, selection and fulfillment reliability.

  Strengths

  Scale, logistics, assortment, pricing and direct commerce capabilities.

  Weaknesses

  Customer-ownership tradeoff and exposure to external-agent access decisions.

  Comparison

  Both are viewed as advantaged over retailers unable to match price and fulfillment variables.

  Risks

  Margin, pricing, macro, regulatory and execution risks.

• Live Nation (LYV) and StubHub (STUB)

  Agentic discovery could improve ticket discovery and sell-through, but effects differ by inventory model.

  Strengths

  Primary ticketing has differentiated inventory; AI can improve personalization and discovery.

  Weaknesses

  Secondary marketplaces have more commoditized inventory and dependence on customer acquisition.

  Comparison

  Primary platforms are viewed more favorably than secondary marketplaces.

  Risks

  Price transparency, service-fee pressure, AI distribution tolls and weaker ancillary monetization.

Key data

• Near-term high-likelihood US spending opportunity$2.6tnConsumer spending identified as likely to lead agentic-commerce adoption.
• High- and medium-likelihood commerce volumeMore than $12.3tnRoughly 60% of current retail and services spend.
• Illustrative eCommerce-growth impactApproximately 1 percentage pointFrom roughly 2% agentic penetration of high-likelihood card-present spending.
• AI use in product discovery~44% of online buyersBain & Co. figure cited by the report.
• Visa/Mastercard eCommerce tokenization>50%Share of eCommerce transactions stated as tokenized.
• Retail media market~$170bn globally in 2025Statista estimate cited by Goldman Sachs, versus $77bn in 2020.
• Muse downloadsOver 2.8 millionSince release, according to Sensor Tower cited by the report.

Impact & implications

The report argues that value should increasingly follow consumer intent. Long-term winners are more likely to combine distribution, consumer trust, merchant participation, payment infrastructure and the ability to convert agent-driven intent into reliable transactions; companies reliant on undifferentiated inventory, paid acquisition or legacy fraud signals may face greater pressure.

Risks

• Consumer adoption may be gradual and uneven across categories, particularly for subjective or high-consequence purchases.
• Intent verification, fraud losses, chargebacks, returns and liability allocation remain unresolved and could limit adoption.
• Merchants may restrict third-party agent access to preserve customer relationships, first-party data, loyalty economics and margins.
• AI platforms may ultimately replace paid search with a new distribution toll through referral fees, commissions or preferred-placement charges.
• Authorized agents may reduce the effectiveness of legacy fraud signals and increase cybersecurity requirements.
• Greater price transparency may commoditize inventory and pressure merchant or marketplace margins.

What to watch

• Progress in liability, authorization, consumer-intent and dispute-resolution frameworks.
• Merchant willingness to expose catalogs, payment flows and transaction infrastructure to third-party agents.
• Adoption of protocols including MCP, UCP, Visa Trusted Agent Protocol and Mastercard Agent Pay.
• Evidence that agents shift product discovery, conversion and eCommerce penetration beyond early use cases.
• Development of AI-native advertising formats, measurement and attribution.
• Changes in fraud intensity, identity-verification demand and payment-network value-added-services attachment.
• Whether AI platforms lower customer-acquisition costs or create new long-term distribution fees.