The most important AI application after ChatGPT! The era of "Online Shopping Agents" is about to arrive. Who will be the winners and who will be the losers?

Author: Bao Yilong

Source: Hard AI

Following ChatGPT, the next major breakthrough in generative AI—the era of "Shopping Agents"—is about to arrive.

On November 17, Morgan Stanley released a research report predicting that by 2030, AI-driven personalized shopping agents could bring an additional $115 billion in incremental spending to the U.S. e-commerce market, accounting for about 6% of total e-commerce spending at that time, and contributing over 100 basis points to annual growth in the industry.

The report emphasizes that not all retailers will benefit. Companies with strong infrastructure, unique inventory, and innovative capabilities, such as Amazon and Walmart, are likely to emerge as winners. In contrast, companies that rely on high commission models, product homogeneity, or are heavily dependent on search traffic, such as Etsy, Chewy, and Lululemon, face challenges.

Additionally, Morgan Stanley believes that traffic entry points will be reshaped. Platforms with significant user reach (such as META and YouTube) will see their value highlighted, while retail media and open web advertising face the risk of being bypassed.

Moreover, the high-profit search advertising model of search giant Google may be impacted by the low-commission agent model, and whether it can successfully convert massive free traffic into paid transactions will be key to its future growth.

What is a "Shopping Agent"?

Since the release of ChatGPT three years ago, investment and innovation in the AI field have been accelerating.

Morgan Stanley's research team believes that from 2025 to 2027, tech giants are expected to invest a cumulative capital expenditure of about $1.7 trillion in data centers. These investments are giving rise to a new generation of generative AI products, with AI shopping agents seen as the next major breakthrough. This "always-on" "Shopping Agent" can:

• Conduct complex product research and price comparisons across platforms and websites.

• Recommend product combinations based on users' personalized needs. For example, putting together a complete "Top Gun" outfit for a Halloween party.

• Automatically track product prices and inventory, and place orders automatically when conditions are met.

• Enable automated, periodic purchasing of items such as fresh groceries.

Currently, platforms and retailers, including Alphabet, OpenAI, Amazon, and Walmart, have begun to launch early versions of "Shopping Agents."

(Major platforms and retailers are launching early versions of "Shopping Agents")

This will transform the e-commerce "shopping funnel" from traditional search, social, and direct access into a new model that is more conversational, personalized, and interactive.

(Shopping agents can help users complete product searches and discoveries more efficiently across multiple shopping platforms by providing relevant background information and personalized services) The "Online Shopping Agent" will further promote the digitalization of consumer wallets by enhancing consumer utility. According to the Morgan Stanley model forecast, by 2030, the "Online Shopping Agent" will bring an incremental spending of $50 billion to $115 billion to the U.S. e-commerce market. Specifically:

• Market Share Forecast: In the baseline scenario, Agent-driven consumption will account for 10% of total e-commerce; in the optimistic scenario, this proportion will reach 20%.

(In the baseline and optimistic scenarios, e-commerce spending will account for approximately 10% and 20% of total spending, respectively)

• Growth Contribution: By 2030, the "Online Shopping Agent" will contribute over 100 basis points (baseline scenario) or 300 basis points (optimistic scenario) of additional growth to the e-commerce industry each year.

• Key Areas: Personalized grocery shopping is seen as a key trigger point. Between 2026 and 2030, groceries and consumer packaged goods (CPG) will contribute 48% (baseline scenario) to 53% (optimistic scenario) of the total Agent shopping volume.

Currently, the e-commerce penetration rate for groceries is only about 12%. Morgan Stanley believes that the multi-step, multi-modal shopping capabilities of AI agents (such as analyzing refrigerator inventory through photos and automatically generating shopping lists) will be an important driver for increasing penetration. After groceries, home goods, personal care, and clothing are the next categories to benefit.

Major Changes in Digital Advertising and Redistribution of Traffic Entry Points

Morgan Stanley believes that the "Online Shopping Agent" will reshape the traffic distribution pattern of digital advertising.

• Value-Enhancing Platforms: Platforms with massive user bases and strong reach capabilities, such as META (Facebook/Instagram), YouTube, and AppLovin, will see their value as brand-building and product discovery tools further enhanced. These platforms are developing tools that allow small and medium-sized enterprises to automatically create and manage entire advertising campaigns through AI Agents.

(META and Google's platforms already have significant advantages in terms of distribution range and coverage capabilities)

• Areas Facing Risks: Retail media and open web advertising will face the greatest risks. As consumers increasingly make shopping decisions through Agents, traffic directly accessing retailer websites will decrease, thereby weakening the value of retail media advertising.

(Currently, user traffic entering e-commerce platforms through search, social networks, and direct access will all be affected by the "Online Shopping Agent")

• Google Faces Transformation: Google Search is at a crossroads. Reports estimate that the "effective commission rate" of Google search ads (the percentage of advertising spending relative to the total gross merchandise volume, GMV) is currently around 33%, while early Agent transaction commissions offered by providers like OpenAI are only in the single digits, a difference of 5 to 10 times. If Google cannot dominate Agent shopping, its high-profit model will be severely eroded.

(The early Agent transaction commissions offered by providers like OpenAI are only in the single digits)

However, the report emphasizes that over 80% of retailers' online traffic on Google is still free, accessible directly and through organic search.

If Google can leverage its Agent products, such as Gemini integrated into the Chrome browser, this could lead to some free traffic converting into paid transactions. Even if the monetization rate per transaction decreases, overall revenue may not be affected and could even grow.

Analysis shows that a 5% shift of free traffic to paid channels can offset a 14% decline in effective monetization rate.

Winners and Losers Among Retailers

Currently, high-margin retail media advertising revenue averages about 6% of e-commerce GMV, often constituting a significant portion or all of the profits for e-commerce platforms.

The research report suggests that when consumers complete purchases through third-party Agents (like OpenAI's ChatGPT), retailers not only lose the opportunity to directly reach customers but may also need to pay commissions to the Agents, which is a double blow to their profit margins.

Morgan Stanley's analysis through the "breakeven curve" indicates that if a third-party Agent charges a 5% commission, retailers need to ensure that about 50% of Agent transactions are "incremental" (i.e., transactions that would not have occurred otherwise) to achieve EBIT (earnings before interest and taxes) breakeven.

(EBIT breakeven points for Agent transaction businesses of various companies)

Additionally, Morgan Stanley proposed a "Five I" framework for assessing retailers' AI agent positioning:

1. Inventory: Unique inventory is becoming increasingly important as agents make cross-platform price comparisons and personalized matching easier.

2. Infrastructure: Logistics infrastructure is key to improving inventory availability and shortening delivery times.

3. Innovation: Investment in GPU-driven machine learning, better matching, and supply management.

4. Incrementality: The ability to drive incremental revenue growth and market share enhancement

5. Income Statement Impact: The risk of agents eroding high-margin core businesses and retail media advertising revenue.

Based on this framework, Amazon, eBay, and Revolve are rated as the most favorably positioned, while Etsy, Chewy, and FIGS face significant challenges. Specifically:

Potential Winners:

• Amazon: With a deep fulfillment network and a Prime membership moat, there are significant opportunities, especially in the fresh grocery sector.

• Walmart: With strong scale, a well-established supply chain network, and a solid membership base, its retail media business faces limited downside risk.

• eBay: With a unique and vast inventory of second-hand and off-season goods, agents can help better connect buyers with long-tail products.

• Other Strong Performers: Companies like Revolve Group, Macy's, and Wayfair also show strong advantages within the framework.

Potential Losers:

• Etsy: Its transaction take rate of over 20% faces immense pressure under the price transparency trend brought by agents.

• FIGS: Highly reliant on search traffic and with high pricing, it is vulnerable to shocks.

• Chewy: Primarily sells homogeneous pet products; although its subscription model is a moat, the products themselves are easily comparable and substitutable.

• Other Disadvantaged Players: Companies like Lululemon, PVH Corp, Dollar Tree, and Five Below are at a disadvantage due to low product differentiation, low digitalization, or excessive reliance on wholesale channels