Agentic commerce presents more opportunity than risk for retailers, with near-term disintermediation risk still low
AI summary card
Agentic commerce presents more opportunity than risk for retailers, with near-term disintermediation risk still low
JPMorgan believes LLMs currently primarily serve as stronger product-discovery entry points, while retailers can retain control of transactions and expand customer acquisition opportunities through UCP, in-platform agents, data, and fulfillment capabilities.
- AI platforms currently contribute less than 1% of traffic to key e-commerce websites, with consumers using them more for product discovery, inspiration, and basket building than for fully automated transactions.
- As retailers move from model-controlled checkout through OpenAI Instant Checkout toward retailer app integration and Google UCP, they are regaining control over transactions, pricing, inventory, membership, and fulfillment.
- WMT is viewed as leading in agentic commerce, followed by TGT; W, BBY, HD, LOW, ULTA, ASO, and others have also partnered with Google UCP.
- Key risks are concentrated in low-engagement, commoditized, highly recurring grocery and CPG categories, as well as strong brands such as Apple and Nike with DTC capabilities that may bypass retail channels.
- The bearish scenario for retail media has moderated because retailers still retain customer data, transaction control, and advertising attribution capabilities.
Report interpretation
Overview
This report discusses the impact of AI and agentic commerce on broadlines, hardlines, and leisure retailers. The core conclusion is that agentic commerce currently functions more as an incremental traffic entry point and product-discovery tool than as an immediate disruption to retailers' transaction and fulfillment positions. Large retailers that connect early to major LLMs, build proprietary platform agents, and strengthen their data infrastructure may instead gain opportunities in customer acquisition, conversion, basket building, and retail media monetization.
Core views
The report believes disintermediation risk remains relatively low because LLMs do not control physical fulfillment networks, retailers have regained control of checkout, and they still own inventory, pricing, membership, fulfillment, and retail media data. Risk is not evenly distributed: grocery and CPG are more exposed because of low engagement, high repeat purchase rates, and strong price discovery; strong brands with DTC capabilities in apparel and footwear may also bypass retailers. By contrast, beauty, pet, home décor, project-based home improvement, and other high-engagement shopping categories face lower risk.
Analysis framework
The report evaluates the impact of agentic commerce across consumer usage stages, transaction control, fulfillment capabilities, data and advertising attribution, brand DTC capabilities, category engagement, and retailer technology investment. It also compares the progress of retailers including WMT, TGT, BBY, HD, LOW, ULTA, ASO, WSM, and DKS in connecting to third-party LLMs and developing proprietary agents.
Methodology notes
Assess the probability of retailers being bypassed based on transaction control, fulfillment control, data control, and brand DTC capabilities.
If AI platforms only handle discovery and basket building while retailers retain control of checkout and fulfillment, disintermediation risk is low; if brands have mature DTC fulfillment and seek to capture retail profits, risk increases.
Low-engagement, commoditized, repeat-purchase products are better suited to automated replenishment, while high-engagement and experiential products depend more on browsing and decision-making.
Grocery and CPG are more susceptible to AI-driven price discovery and automated replenishment, while beauty, pet, and home décor depend more on experience, emotion, and basket building and therefore face lower risk.
Retailers should both connect to external LLMs such as ChatGPT and Gemini and build agents for their own apps and websites.
External LLMs bring customer acquisition and discovery traffic, while proprietary platform agents help retain customer relationships, improve conversion, collect proprietary data, and support retail media.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- WMTLeading beneficiary
- Strengths
- Broad category breadth, strong marketplace capabilities, large-scale fulfillment, and an early launch of Sparky embedded in its own app and third-party LLM scenarios.
- Weaknesses
- High grocery exposure creates long-term vulnerability to automated replenishment and price discovery pressure.
- Comparison
- The report considers WMT the most advanced retailer, followed by TGT.
- Risks
- If LLMs eventually control the transaction entry point and compress retail media, grocery-related margins could come under pressure.
- TGTFollow-on beneficiary
- Strengths
- Has built in-app experiences for ChatGPT and Google Gemini supporting search, basket building, and final checkout.
- Weaknesses
- Its proprietary platform agent remains in testing and refinement across different stages.
- Comparison
- Behind WMT but ahead of most other retailers.
- Risks
- If data connectivity, inventory accuracy, or payment experiences are inadequate, users may remain at the information-search stage rather than completing transactions.
- BBYCategory data beneficiary with brand DTC risk
- Strengths
- Has deep expertise and product-matching data in electronics, helping LLMs map complex needs to suitable products.
- Weaknesses
- Has historically been affected by channel disintermediation by major electronics brands.
- Comparison
- Complex electronics have greater advisory needs than ordinary products, but single-brand products may be easier to bypass.
- Risks
- Brands such as Apple with DTC capabilities may bypass retailers, while greater price transparency enabled by LLMs may compress margins.
- HD / LOWRelatively low risk with proprietary agent opportunities
- Strengths
- Project-based home improvement has high fulfillment complexity and demand for professional advice; HD's Magic Apron and LOW's MyLow can be used for recommendations and basket building.
- Weaknesses
- They need to maintain accurate inventory, store-level availability, and delivery commitments.
- Comparison
- Compared with grocery and standardized consumer products, high-engagement home-improvement shopping is less likely to be fully automated.
- Risks
- If external LLM information does not match actual prices, inventory, or delivery speed, conversion could suffer.
- ULTALow-risk experiential retailer
- Strengths
- Beauty is an emotion- and experience-driven category; membership and customer data are highly valuable, and ULTA AI can enhance personalized recommendations.
- Weaknesses
- It needs to continue improving its proprietary data infrastructure and LLM connectivity.
- Comparison
- Compared with commoditized categories, beauty shopping depends more on discovery, trial, and preferences.
- Risks
- If brands strengthen DTC or AI platforms capture the discovery entry point, some traffic could be weakened.
- WSMDefensive beneficiary of brand ownership
- Strengths
- The report states that nearly 90% of sales are driven by proprietary brands, which can mitigate the risk of bypass by strong third-party brands with DTC capabilities.
- Weaknesses
- Home furnishings still require excellent browsing, design advice, and delivery experiences.
- Comparison
- Its proprietary-brand structure is more favorable than that of retailers dependent on strong third-party brands.
- Risks
- If proprietary agents and external LLM connectivity are insufficient, the company may miss high-intent traffic.
- DKSMore exposed to brand behavior
- Strengths
- Sporting goods have some experiential and specialized shopping attributes, and DKS has already tested proprietary agents such as Coach.
- Weaknesses
- Some core brands have strong DTC capabilities.
- Comparison
- Nike has historically bypassed retail partners at certain stages, and the report uses it as a risk case.
- Risks
- If brands such as Nike use agentic commerce to gain market share and capture retail profits, they could reduce the role of retailers.
Key data
- AI Platform Traffic Share<1%The report states that AI platforms still account for less than 1% of total traffic across key e-commerce markets and retail websites.
- Share of ChatGPT Queries for Purchasable ProductsApproximately 2%Citing a September 2025 NBER study, the report states that approximately 2% of ChatGPT queries relate to purchasable products.
- Share of Information-Seeking QueriesApproximately 21%The same study shows that approximately 21% of queries are used to seek information.
- Share of Users Using AI for Purchase DecisionsApproximately 43%A JPMorgan private-company research consumer survey shows that approximately 43% of users have used AI to make purchase decisions.
- Share of Users Completing Transactions via AI ChatbotsApproximately 9%The survey shows that the proportion of users actually completing transactions through AI chatbots remains low.
- Early OpenAI Instant Checkout FeeApproximately 4%The report states that OpenAI initially charged retailers an average affiliate fee of approximately 4% on transactions completed through Instant Checkout, although this model has weakened.
- Google UCP Partnership ParticipantsApproximately 20 companiesThe report states that approximately 20 retailers and internet companies participate in the Google UCP council.
- WMT Sparky UsageUsed by approximately half of app users, with basket sizes 35% higherThe report states that WMT disclosed that approximately half of its app users use Sparky and that basket sizes are 35% higher.
Impact & implications
For retail stocks, agentic commerce is more likely to reinforce near-term advantages for the strongest players: large retailers with scale, fulfillment, membership data, inventory visibility, and advertising technology are best positioned to convert LLM traffic into sales and data assets. WMT, TGT, and retailers that have connected to Google UCP or are building proprietary agents are better positioned; grocery, CPG channels, and retailers reliant on sales of strong third-party brands should remain alert to price discovery, channel disintermediation, and advertising revenue shifts.
Risks
- LLMs' enhanced price discovery capabilities could intensify competition in grocery, CPG, and other price-sensitive categories.
- If AI agents become sufficiently mature to control complete transactions and checkout, retailers could lose transaction data and high-margin advertising revenue.
- Strong brands with DTC fulfillment capabilities may bypass retailers, particularly in apparel, footwear, and certain consumer electronics products.
- Inconsistencies between external AI interfaces and retailers' actual prices, promotions, membership benefits, inventory, and delivery information could reduce consumer trust and conversion.
- Retailers that fail to connect promptly to major LLMs or build proprietary agents may lose traffic and customer data through new entry points.
What to watch
- User penetration, conversion rates, and basket changes for proprietary agents such as WMT Sparky, the TGT app experience, HD Magic Apron, LOW MyLow, ULTA AI, WSM Olive, ASO Scout, and DKS Coach.
- The implementation speed of Google UCP, OpenAI ACP, and retailer app integrations, and whether checkout control remains with retailers.
- Whether AI platform traffic rises significantly from its current level of less than 1% of total e-commerce website traffic.
- Whether consumers move beyond product discovery and information searches toward completing transactions and automated replenishment through AI.
- Whether retail media revenue is diverted to AI platform advertising or benefits from stronger customer data and contextual advertising conversion.
- Changes in price competition and channel strategies in high-risk categories such as grocery, CPG, branded apparel and footwear, and consumer electronics.