At present, opportunities of Agentic Commerce outweigh the risks, and retailers’ disintermediation risk remains low

Overview

This report discusses the impact of AI and Agentic Commerce on the broadlines and hardlines retail sector. The core conclusion is that LLMs currently mainly serve as an entry point for product discovery, information lookup, and basket building, acting more like a stronger version of search rather than a role already capable of broadly substituting retailers’ transaction and fulfillment functions. Because retailers still control checkout, customer relationships, inventory information, and fulfillment, the report argues that disintermediation risk remains low and Agentic Commerce is more of an opportunity for large retailers.

Core views

Key points include: first, AI platforms currently generate less than 1% of e-commerce traffic and the monetization path is still unclear, with AI platforms shifting focus from direct Agentic transactions to digital advertising and app integration. Second, retailers need to engage with LLMs such as ChatGPT and Gemini early, while also building proprietary platform Agents to enhance acquisition, retention, product discovery, basket building, and proprietary data accumulation. Third, transaction control is moving from model-controlled environments like OpenAI Instant Checkout toward retailer application integration and UCP-style retailer-controlled environments, which is critical to reducing retail media and customer data leakage risk. Fourth, category risk differs materially: low-involvement, standardized, high-repurchase-giving grocery/CPG categories carry higher risk, while beauty, pet, and home decor with emotional and browse-driven attributes have lower risk.

Analysis framework

The report uses an industry thematic analysis framework, breaking Agentic Commerce into dimensions including traffic ingress, transaction control, fulfillment capability, customer data, retail media, and category characteristics, and performs a cross-sectional comparison using the LLM partnership and proprietary Agent progress of companies including WMT, TGT, BBY, HD, LOW, ULTA, ASO, W, DKS, and WSM.

Methodology notes

• Industry structure analysisDisintermediation risk framework

  Transaction control, fulfillment control, customer data control

  The report argues that disintermediation risk rises materially only when LLMs control both transactions and the data and ad-revenue levers; at present, checkout and fulfillment remain mainly managed by retailers.

• Category risk analysisInvolvement and substitutability framework

  Low involvement, repurchase frequency, standardization, DTC capability

  Grocery/CPG carries higher risk because price discovery, repurchase behavior, and product standardization are stronger; categories with higher involvement or experience-driven demand face lower risk because consumers still value browsing and choice.

• Competitive strategy analysisDual-track strategy of LLM partnerships and proprietary Agents

  Parallel integration of third-party LLMs and proprietary platform Agents

  The report recommends retailers connect to major LLMs to capture incremental traffic while also developing proprietary Agents to accumulate data, improve conversion, and enhance customer support.

Asset mapping & comparison

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

• WMT

  Leading beneficiary

  Strengths

  Sparky is most advanced, product coverage is broad, market platform and fulfillment capabilities are strong, and shopping assistant capabilities are advancing in both third-party LLMs and its own app.

  Weaknesses

  Its grocery exposure is relatively high, with long-term pressure on margins from price discovery and automated replenishment.

  Comparison

  The report views WMT as a clear leader among retailers, with TGT in second position and other firms following quickly.

  Risks

  If LLMs eventually control transactions and weaken retail media touchpoints, grocery and advertising income could come under pressure.

• TGT

  Follower beneficiary

  Strengths

  Has built app experiences in ChatGPT and Google Gemini, supporting search, basket construction, and checkout.

  Weaknesses

  Its proprietary platform Agent remains at varying beta stages and is less mature than WMT.

  Comparison

  It is behind WMT and is viewed in the report as a retailer taking a relatively proactive stance on Agentic Commerce.

  Risks

  If pricing, inventory, or payment experiences are inconsistent, consumers may remain at the information-search stage rather than complete purchases.

• BBY

  Lower to moderate risk retailer

  Strengths

  Has deep product expertise and data in complex categories, helping match the best products for specific usage scenarios.

  Weaknesses

  Needs better integration between site data and LLMs.

  Comparison

  Its electronics category has been pressured by AMZN, but complex-product matching capability can be a source of advantage in the LLM era.

  Risks

  Strong brands such as Apple could weaken retailer channel value if they strengthen DTC.

• HD/LOW

  Relatively low risk with AI service upside

  Strengths

  Home-improvement categories generally have high involvement and strong needs for service and recommendations; proprietary Agents like Magic Apron and MyLow can improve shopping and customer support experience.

  Weaknesses

  Need to ensure local information on inventory, pricing, and delivery speed is consistent across external LLMs and proprietary platforms.

  Comparison

  Compared with low-involvement products, home categories rely more on project-based buying and professional advice.

  Risks

  If Agent recommendations become overly commercialized, user experience and conversion may suffer.

• ULTA

  Relatively low-risk category representative

  Strengths

  Beauty is an emotional and high-involvement category with high loyalty and customer data value, and ULTA AI can strengthen personalized recommendations.

  Weaknesses

  Needs ongoing investment in data and Agent experience.

  Comparison

  The report classifies beauty as a lower-risk basket, unlike grocery-style high-risk categories.

  Risks

  Migration of brand DTC and ad budgets still needs monitoring.

• Grocery/CPG

  High-risk category

  Strengths

  High frequency, repurchase behavior, and automated replenishment characteristics make it possible for Agentic Commerce to raise online penetration.

  Weaknesses

  Low involvement, product standardization, and stronger price discovery could compress retailer gross margins.

  Comparison

  Compared with browse-driven categories such as beauty, pet, and home decor, grocery is more likely to be reshaped by automated purchasing flows.

  Risks

  If LLMs control transactions but leave fulfillment to retailers, retailers may lose high-margin advertising revenue.

• Strong DTC brands like NKE/Apple

  Potential source of disintermediation

  Strengths

  High brand recognition and direct-to-consumer channels with fulfillment capability.

  Weaknesses

  If they increasingly bypass retailers, channel management and distribution relationships can be affected.

  Comparison

  The report cites Nike’s historical disintermediation of retail partners such as DKS as a reference case.

  Risks

  If they capture retail profits through Agentic Commerce, relevant retailers may lose traffic and profit pools.

Key data

• Share of e-commerce traffic from AI platforms<1%The share of AI platforms in total traffic on key e-commerce markets and retailer sites remains low.
• Share of ChatGPT shopping-related queries that are purchase-relatedabout 2%The report cites a 2025 September NBER study showing purchase-related queries remain a minority.
• Share of information queriesabout 21%The same study shows that many users still mainly use AI for information search.
• Share of users using AI in purchase decisionsabout 43%A JPM Private Company Research consumer survey indicates AI is already influencing purchase decisions.
• Share of users completing transactions with AI chatbotsabout 9%Completion conversion remains materially lower than decision-assistance usage.
• Early merchant fees for OpenAI Instant Checkoutabout 4% on averageThe report says OpenAI previously charged retailers an affiliate fee for transactions completed through Instant Checkout, and that model has since been deemphasized.
• Number of parties participating in Google UCPabout 20 companiesIncludes AMZN, BBY, HD, LOW, META, MSFT, ETSY, SHOP, TGT, ULTA, WMT, and W.
• WMT Sparky usage and basket performanceabout half of app users; basket size is up 35%The report says WMT disclosed high Sparky engagement and a larger average basket size.
• WSM private label sales sharenearly 90%A high private-label share reduces risk of external brand DTC disintermediation.
• J.P. Morgan global equity research rating mixOverweight 53%, Neutral 36%, Underweight 12%As of the disclosure on 2026-07-04, rounding may cause the total not to equal 100%.

Impact & implications

For investment judgment, this implies that market concerns about AI platforms disintermediating retailers may be overstated. In the near term, the more important question is who can most quickly integrate into the LLM ecosystem while retaining transaction and data control, improve local inventory and pricing accuracy, and convert AI traffic into higher-intent customers. Large retailers are more likely to benefit due to scale, fulfillment networks, customer data, and existing ad business, while grocery, CPG, DTC brand-led apparel and footwear, and some consumer electronics still need monitoring for pressure on margin from price transparency and brand bypass channels.

Risks

• If AI platforms again gain control over transactions and checkout in the future, retailer customer data, advertising income, and channel control could be weakened.
• Grocery/CPG may face stronger margin pressure because of price transparency, repurchase behavior, and low involvement characteristics.
• Strong brands with mature DTC fulfillment capabilities may bypass retailers, particularly in apparel, footwear, and some consumer electronics categories.
• If external Agent interfaces are inconsistent with retailer sites on price, inventory, promotions, loyalty, or delivery speed, consumer adoption may be limited.
• If retailers over-promote in AI recommendations or load AI experiences with too much advertising, user experience and trust may be damaged.

What to watch

• User engagement, conversion rates, and basket size of WMT Sparky, TGT App in ChatGPT/Gemini, and proprietary Agents from other retailers.
• Adoption pace of protocols in e-commerce contexts such as Google UCP, OpenAI ACP, Stripe checkout, and Anthropic MCP.
• Changes in the share of AI platform traffic in total e-commerce site traffic, especially whether it rises materially above the below-1% level.
• Whether consumers shift from using AI for discovery and basket building to completing transactions within AI platforms.
• Growth in retail media revenue, ad conversion rates, and whether brand ad budgets shift toward AI platforms.
• Signs of DTC brands in grocery, CPG, footwear/apparel, and consumer electronics bypassing retailers.
• Accuracy of retailer local inventory, pricing, promotion, and loyalty data in interfaces with LLMs.