All reports
Trends by deep-research

Agent-Mediated Commerce — From AI-Assisted Discovery to AI-Executed Checkout

OctantOSAgentScope

Agent-Mediated Commerce — From AI-Assisted Discovery to AI-Executed Checkout

Research date: 2026-03-19 | Agent: Deep Research | Confidence: High

Executive Summary

  • Agent-mediated commerce is moving from experimentation to execution: Adobe reports +693.4% YoY AI-driven retail traffic in holiday 2025, with AI-referred sessions converting 31% higher than non-AI traffic.
  • The funnel is collapsing into chat interfaces: Walmart + OpenAI and Google + major retailers (Walmart, Shopify, Wayfair) are pushing in-chat instant checkout flows.
  • Adoption is rising, but trust is the bottleneck: Capgemini shows 25% of consumers already used GenAI shopping tools in 2025, yet 76% want strict control rules and 71% worry about data usage.
  • Market upside is meaningful: Morgan Stanley estimates $190B-$385B U.S. e-commerce spend by 2030 via agentic shoppers (10-20% share), while McKinsey estimates up to $900B U.S. B2C retail revenue could be orchestrated by agentic commerce by 2030.
  • Go/No-Go for Moklabs: Go for B2B infra (trust/governance/merchant-readiness/orchestration tooling). No-Go for a direct-to-consumer shopping agent competing with OpenAI/Google/Amazon.

Market Size & Growth

TAM, SAM, SOM (with methodology)

LayerEstimateMethodologyConfidence
TAMUp to $900B U.S. B2C retail orchestrated by 2030McKinsey estimate for agentic-commerce orchestrated revenue in U.S. B2C retail by 2030.Medium-High
SAM$190B-$385B U.S. e-commerce by 2030Morgan Stanley estimate for agentic shoppers’ share (10-20%) of U.S. e-commerce.High
SOM$4M-$15M annual revenue potentialAssumption: Moklabs infrastructure captures 0.1% of SAM GMV and monetizes at 2-4% blended software + infra take rate.Medium

Growth Signals

  • Adobe (Jan 2026): AI traffic to retail sites grew 693.4% YoY; U.S. online holiday spend reached $257.8B.
  • Adobe (Jan 2026): AI referrals converted 31% higher than other channels; AI revenue-per-visit was +254% YoY holiday season-to-date.
  • Amazon AWS (Nov 2025): Rufus reached 250M+ users, MAU up 140% YoY, interactions up 210% YoY, with users 60% more likely to complete purchase.
  • AP (Jan 2026): Salesforce estimate cited in coverage says AI influenced $272B (20%) of global holiday retail sales.

Key Players

CompanyFoundedFundingRevenue/ARRPricingKey Differentiator
OpenAI + Walmart2015 / 1962Private + PublicWalmart FY2025 revenue $681BNot publicly disclosed for Instant CheckoutChat-native checkout connected to Walmart logistics and catalog
Google (Gemini + Shopping + UCP)1998PublicProduct-level revenue not disclosedConsumer-facing features bundled; merchant terms varyShopping Graph scale (50B+ listings, 2B hourly refresh) + UCP standard push
Amazon Rufus1994PublicProduct-level revenue not disclosedIncluded in Amazon shopping experienceMassive user base, high-intent conversion uplift, deep catalog integration
Visa Intelligent Commerce + TAP1958PublicProduct-level revenue not disclosedEnterprise/developer partnership modelTrusted Agent Protocol and payment-network-grade risk/auth controls
Stripe ACP Suite2010PrivateProduct-level revenue not disclosedDeveloper-led; pricing varies by Stripe products usedExplicit agentic checkout protocol (ACP) + programmable payment flows
Wayfair (UCP co-developer)2002PublicProduct-level revenue not disclosedMerchant-side commerce modelMerchant-of-record-preserving UCP checkout integration in Google surfaces

Technology Landscape

Dominant Stack for Agent-Mediated Commerce

  1. Discovery and relevance layer
  • LLM/chat interfaces (ChatGPT, Gemini, Rufus)
  • Large product graphs/catalogs (Google Shopping Graph)
  1. Merchant data layer
  • Structured product feeds
  • Inventory, price, shipping, and policy metadata
  1. Protocol + orchestration layer
  • UCP (Google ecosystem), ACP (Stripe), TAP (Visa)
  • Agent workflow orchestration with explicit user intent checkpoints
  1. Checkout and payment layer
  • Tokenized credentials
  • Agent-authenticated payment authorization
  • Human confirmation step for high-risk transactions
  1. Trust and governance layer
  • Bot/agent identity verification
  • Fraud detection and abuse controls
  • Audit trails and policy enforcement
  • From AI-assisted search to AI-executed purchase: checkout is moving into the same interface as discovery.
  • Standard wars are starting: UCP, ACP, and payment-network protocols are converging but not unified.
  • Merchant-of-record is strategic: retailers want agentic distribution without giving up ownership of customer and fulfillment relationship.

Open Source vs Proprietary Dynamics

  • Open protocols are accelerating integration but remain fragmented.
  • Proprietary platforms still control demand aggregation and consumer attention.
  • The infrastructure moat is in compliance, trust scoring, and operability across standards.

Pain Points & Gaps

Unmet Needs

  1. Trust/consent controls are underdeveloped relative to execution capability.
  2. Protocol fragmentation (UCP vs ACP vs proprietary APIs) creates integration tax.
  3. Merchant readiness is low: most stores are not exposed in machine-readable, agent-ready formats.
  4. Cross-platform attribution is weak: difficult to measure ROI per agent channel vs traditional paid/organic channels.
  5. Post-purchase reliability (returns, disputes, support handoff) is still immature in agentic flows.

Common Complaints (Reddit, HN, Industry Threads)

  • Merchant confusion on standards and onboarding (“which protocol actually matters first?”).
  • Skepticism about checkout value-add versus existing autofill/payment rails (HN discussion on Instant Checkout).
  • Support loops and poor escalation when AI-mediated support fails (recurring Shopify community/Reddit pattern).
  • Concern that over-automation reduces user control in higher-stakes purchases.

Opportunities for Moklabs

Ranked Opportunities (Effort/Impact)

OpportunityEffortImpactTime-to-marketResource NeedConnection to Moklabs
1) Agentic Commerce Readiness Scanner (UCP/ACP/TAP checks)Low-MediumHigh3-5 weeks2 engineersFits AgentScope observability and QA-style diagnostics
2) Trust & Governance Layer for Agent CheckoutMediumVery High6-9 weeks3 engineers + 1 securityExtends Argus security + Paperclip governance DNA
3) Merchant Adapter SDK (Shopify/WooCommerce/Custom)Medium-HighHigh8-12 weeks4 engineersAligns with Octant/AgentScope interoperability thesis
4) Post-Purchase Agent Orchestrator (returns/disputes/support)MediumMedium-High6-8 weeks3 engineersOperational wedge with clear ROI and less platform conflict
5) ROI Analytics for Agentic FunnelMediumHigh5-7 weeks2 engineers + 1 analystNatural extension of prior ROI/cost attribution research

Go/No-Go Recommendation

  • GO: Build B2B infrastructure products that help retailers become agent-ready and compliant across multiple ecosystems.
  • NO-GO: Build a consumer shopping assistant competing for consumer surface area against OpenAI, Google, Amazon, and payment networks.
  • Why: Distribution power is concentrated, but governance/interoperability tooling remains fragmented and under-served.

Risk Assessment

Market Risks

RiskLikelihoodImpactMitigation
Big platforms internalize infra capabilitiesHighHighFocus on cross-platform neutrality and compliance differentiation
Slower consumer trust adoptionMediumMedium-HighLead with merchant pain (ops, fraud, attribution), not consumer novelty
Protocol fragmentation persistsHighMediumBuild adapter-first architecture; avoid single-protocol lock-in

Technical Risks

RiskLikelihoodImpactMitigation
Protocol churn and backward incompatibilityHighHighVersioned adapters + conformance test suite
Fraud and abuse escalation via autonomous agentsMedium-HighHighIdentity + behavioral verification + HITL for risky transactions
Incomplete observability across channelsHighMediumUnified event schema and trace IDs from discovery to payment

Business Risks

RiskLikelihoodImpactMitigation
Difficult enterprise sales motion in crowded AI marketMediumHighStart with diagnostic/readiness product as land strategy
Liability around failed purchases/disputesMediumHighKeep merchants as record owners; provide policy/audit layer only
Pricing pressure from bundled hyperscaler offeringsMediumMediumMonetize advanced governance and multi-network optimization

Data Points & Numbers

Data PointValueSourceConfidence
AI-driven traffic growth to retail sites (holiday 2025)693.4% YoYAdobe Newsroom (Jan 2026)High
U.S. online holiday spend (Nov-Dec 2025)$257.8BAdobe Newsroom (Jan 2026)High
AI referrals conversion lift vs non-AI+31%Adobe for Business blog (Jan 2026)High
AI revenue-per-visit lift+254%Adobe for Business blog (Jan 2026)Medium-High
Consumers who trust AI in shopping (survey)47%Adobe for Business blog (Jan 2026)Medium
Rufus users in 2025250M+AWS ML Blog (Nov 2025)High
Rufus monthly users growth+140% YoYAWS ML Blog (Nov 2025)High
Rufus interaction growth+210% YoYAWS ML Blog (Nov 2025)High
Purchase likelihood with Rufus60% more likelyAWS ML Blog (Nov 2025)Medium-High
Shopping Graph catalog size50B+ listingsGoogle Shopping blog (Nov 2025)High
Shopping Graph update frequency2B listings updated hourlyGoogle Shopping blog (Nov 2025)High
Capgemini: GenAI shopping tool use in 202525% of consumersCapgemini Consumer 2026High
Capgemini: plan to use GenAI shopping tools31%Capgemini Consumer 2026High
Capgemini: want clear AI action rules76%Capgemini Consumer 2026High
Capgemini: concerned about data usage71%Capgemini Consumer 2026High
Morgan Stanley agentic spend estimate (US, 2030)$190B-$385BMorgan Stanley ResearchHigh
Morgan Stanley estimated market share (US e-commerce, 2030)10%-20%Morgan Stanley ResearchHigh
Americans purchasing via AI in prior month23%Morgan Stanley ResearchMedium
McKinsey orchestrated revenue opportunity (US B2C, 2030)Up to $900BMcKinsey Agentic Commerce reportMedium-High
AP-cited Salesforce estimate of AI-influenced holiday sales$272B (20% global retail sales)AP News coverageMedium
Walmart FY2025 revenue$681BWalmart announcementHigh

Sources

Related Reports