Executive Summary

The global retail ecosystem stands at the precipice of a structural transformation that rivals the introduction of the internet itself. For decades, the fundamental mechanics of commerce have remained tethered to human-centric discovery: the browse, the search query, the filter, and the visual assessment of an assortment. Whether in a physical department store or a digital marketplace, the burden of discovery and decision-making has rested firmly on the cognitive load of the consumer.

This paradigm is now dissolving. We are transitioning from an era of “search-based” commerce to “agentic commerce,” a regime where autonomous Artificial Intelligence (AI) agents—powered by Large Language Models (LLMs) and sophisticated reasoning engines—execute discovery, negotiation, and purchasing functions on behalf of the user.

This report provides an exhaustive analysis of this shift, specifically examining the friction between traditional merchandising tenets—the interplay of “deal hunting” and “assortment optimization”—and the emerging logic of algorithmic purchasing. The research indicates that while agentic commerce promises unprecedented efficiency and the realization of “just-in-time” (JIT) consumer utility, it poses an existential threat to the serendipitous discovery mechanisms that have historically driven high-margin impulse purchases, brand affinity, and the emotional gratification of shopping.

The analysis draws upon extensive market data, technical architectural reviews of emerging protocols like the Model Context Protocol (MCP), and behavioral studies of holiday shopping trends in 2025. It argues that retail is bifurcating into two distinct modalities: a hyper-efficient, agent-mediated commodity layer where price and structured data reign supreme, and a hyper-experiential, human-centric layer where “soft attributes” and brand storytelling must be sufficiently potent to bypass the filter bubble of the machine.

The following sections detail the psychological foundations of traditional merchandising, the disruptive mechanics of autonomous agents, the specific impacts on high-stakes periods like the holiday season, and the strategic imperatives for retailers attempting to survive the commoditization of the “zero-click” economy.

Part I: The Foundations of Merchandising and Consumer Psychology

To fully comprehend the disruptive magnitude of agentic shopping, one must first establish the baseline of traditional merchandising principles that have governed retail for over a century. Retail success has historically relied on balancing two distinct, often contradictory, psychological drivers: the rational pursuit of value (the “deal”) and the emotional satisfaction of discovery (the “assortment”). This dialectic forms the core of the “4 Ps” marketing mix—Product, Price, Place, and Promotion—and dictates how physical and digital shelves are constructed.

1.1 The Dialectic of Deal vs. Assortment

The tension between “deal shopping” and “assortment planning” is not merely operational; it is deeply psychological. These two modes represent fundamentally different cognitive states for the consumer, requiring distinct merchandising strategies.

Deal Shopping: The Adversarial Transaction

Deal shopping is inherently transactional and often adversarial. It represents a “hunt” for value where the consumer attempts to maximize utility while minimizing expenditure. In this mode, the consumer is analytical, price-sensitive, and focused on a known quantity or category.

Merchandising strategies targeting this behavior rely heavily on pricing mechanics—markdowns, “loss leaders,” and promotional signaling (e.g., red shelf tags, “slash-through” pricing on web pages).

While effective for driving volume and clearing stagnant inventory, a reliance on deal-centric merchandising carries significant long-term risks. Frequent markdowns can erode brand equity, training consumers to devalue the product and wait for the inevitable sale, thereby compressing profit margins. The “deal” is a blunt instrument; it drives conversion through economic coercion rather than desire.

Assortment Planning: The Art of Curation

Conversely, assortment planning is the art of curation. It is designed to maximize revenue potential not by lowering the price, but by optimizing the product mix to resonate with latent consumer demand and identity.

Strategic assortment planning enhances cash flow and profitability by aligning inventory with the customer’s lifestyle, thereby reducing the need for markdowns.

More importantly, the assortment is the brand. As noted in industry analyses, a carefully curated assortment creates a distinct shopping environment that builds trust and recognition. When a retailer effectively plans an assortment, they are not just stocking shelves; they are constructing a narrative. This consistency reinforces the brand’s image and helps distinguish it from competitors in a crowded marketplace.

Table 1: The Psychological Divergence of Shopping Modes

Feature	Deal Mode (Transactional)	Assortment Mode (Experiential)
Primary Driver	Economic Utility, Savings	Identity Expression, Discovery
Cognitive State	Analytical, High Focus	Relaxed, Open to Stimuli
Merchandising Trigger	Price, Urgency (“Limited Time”)	Visuals, Storytelling, Adjacency
Inventory Role	Liquidation, Volume Driver	Brand Builder, Margin Driver
Discovery Mechanism	Search, Filter by Price	Browse, Serendipity, “Window Shopping”

1.2 The Psychology of Serendipity and the “Thrill of the Hunt”

The bridge between a browser and a buyer is often “serendipity”—the happy accident of finding something unexpected but relevant. Serendipity in retail is rarely accidental; it is engineered through rigorous visual merchandising and store layout principles.

Imagine walking into a favorite clothing store and immediately spotting a “perfect shirt” that was not on the shopping list. This discovery is the result of carefully applied merchandising principles: color blocking, complementary product adjacency, and visual hierarchy. This “thrill of the hunt” is a powerful dopaminergic driver. Research indicates that visual merchandising and product variety significantly impact impulse buying behavior, driven by hedonic shopping motivations.

The Role of “Window Shopping”

The concept of “window shopping”—browsing without immediate intent—is critical for the absorption of “soft attributes.” These are the intangible qualities of a brand or product: its “vibe,” heritage, aesthetic signaling, and cultural relevance. In a physical environment, or a rich digital storefront, a consumer absorbs these cues passively. They might enter a luxury boutique not to buy, but to participate in the brand’s atmosphere, which builds long-term affinity.

Digital merchandising has attempted to replicate this through recommender systems (“People who bought X also bought Y”). However, traditional recommender systems function as assistants to a human decision-maker. They suggest; they do not decide. The human remains the final arbiter, retaining the ability to be distracted, delighted, or swayed by an irrational emotional appeal. This dynamic is fundamental to the high margins of the fashion, beauty, and lifestyle sectors, where purchase decisions are rarely strictly rational.

1.3 The First Principles of Selling: Trust and Experience

Underpinning all merchandising is the principle of trust. The retailer exists for the customer, and successful selling requires a deep understanding of target demographics. Trust is built through consistency—ensuring that the brand experience (BX) is uniform across all touchpoints.

In the traditional model, the brand owns the experience. From the packaging design (the “unboxing experience” pioneered by brands like Apple) to the store layout, every element is a controlled variable designed to foster loyalty. Experience design (XD) aims to make the journey seamless and intuitive, but crucially, it is a journey that the customer takes. The customer is the traveler; the brand is the guide. As we will see, agentic commerce threatens to displace the traveler entirely, leaving the brand guiding a machine that does not care about the scenery.

Part II: The Rise of the Agentic Consumer

We are currently witnessing a phase shift from intent-based search to delegated execution. This is the dawn of the “Agentic Consumer.” Agentic commerce involves software systems, typically powered by Generative AI and Large Language Models (LLMs), that do not merely answer questions but act on the user’s behalf to browse, compare, negotiate, and initiate purchases.

2.1 Defining Agentic Commerce and Autonomous Agents

Unlike traditional chatbots or search engines, AI agents possess “agency”—the ability to pursue a goal autonomously across multiple steps. As outlined by McKinsey, these agents anticipate needs, navigate shopping options, and execute transactions, effectively “riding the rails” of existing digital commerce infrastructure.

The market potential is staggering. Projections suggest that by 2030, the U.S. B2C retail market alone could see up to $1 trillion in orchestrated revenue from agentic commerce, with global impact reaching $3-5 trillion. This is not a future scenario; it is an unfolding reality. In 2024, retailers saw a 1,200% surge in traffic originating from GenAI platforms, signaling that the shift is already underway.

The Functional Mechanics of Agents

Agents operate differently from human shoppers. They prioritize structured data over visual appeal. According to analysis by BCG and Kearney, agents prioritize price, user ratings, delivery speed, and real-time inventory availability over brand familiarity or loyalty. They are “rational actors” in the economic sense, immune to the emotional manipulation of traditional advertising.

Goal-Directed Behavior: An agent is given a directive (e.g., “Find a waterproof hiking boot under $150 with 4+ stars delivered by Friday”). It will relentlessly filter the entire web to satisfy these constraints.
Multi-Step Reasoning: Agents can chain tasks. They can read a review, cross-reference it with a return policy, check a third-party shipping estimate, and then execute the purchase.
Context Awareness: Through protocols like the Model Context Protocol (MCP), agents can maintain “memory” of user preferences, sizes, and past purchases, allowing for hyper-personalized filtering.

2.2 The “Zero-Click” and Anticipatory Commerce Paradigm

The logical conclusion of agentic commerce is “zero-click” shopping. This paradigm envisions a future where the friction of manual website navigation is eliminated entirely. Instead of a user clicking through categories, the transaction occurs in the background, authorized by high-level intent.

This enables “Anticipatory Commerce,” where the system predicts a need before the user explicitly articulates it. AI agents can analyze consumption rates (e.g., predicting when laundry detergent will run out) or calendar events (e.g., an upcoming birthday) and proactively suggest or execute a purchase. This shifts the retail model from a “pull” mechanic (user searching for goods) to a “push” mechanic (goods finding the user).

In this environment, the primary competitive metric shifts. Brands have spent decades fighting for “Share of Attention” via billboards, banner ads, and influencers. In the agentic era, they must fight for “Share of Algorithm”. If an agent utilizes a “best fit” logic, it will filter out products that do not meet strict quantitative criteria before the human ever sees them. For example, Amazon’s “Rufus” agent and similar tools are programmed to find the optimal product, potentially switching brands if a competitor offers a better value proposition or logistical reliability.

2.3 Technical Architecture: How Agents “Think” and “Shop”

To understand the behavior of these agents, one must understand their architecture. They are not simply “Googling” products. They are interacting with the web via specific protocols and data structures.

The Model Context Protocol (MCP)

A critical enabler of this ecosystem is the Model Context Protocol (MCP). MCP allows AI models to securely connect to data sources (like a retailer’s inventory database) and tools (like a payment gateway). MCP standardizes how agents interact with the world, replacing fragmented API integrations with a unified language for commerce. Retailers who adopt MCP servers position themselves to be “readable” by these agents.

Agent-to-Agent (A2A) Communication

The future landscape will likely involve “swarms” of agents communicating via Agent-to-Agent (A2A) protocols. A user’s “Personal Shopper Agent” might negotiate with a retailer’s “Sales Agent.”

Scenario: A consumer’s agent requests a discount on a bulk order of winter coats. The retailer’s agent analyzes the inventory levels, current demand, and the consumer’s “lifetime value” score, and autonomously approves a 10% discount to close the deal. This negotiation happens in milliseconds, without human intervention, fundamentally altering the “fixed price” nature of B2C retail.

Table 2: The Shift from Human-Centric to Agentic Commerce

Feature	Human-Centric Commerce	Agentic Commerce
Discovery Mode	Visual Browsing, Serendipity	Parametric Filtering, Prediction
Primary Driver	Emotion, Brand Affinity, Impulse	Logic, Efficiency, “Best Fit”
Brand Loyalty	Sticky (Habit, Identity)	Fluid (Performance-based)
Sensitivity	High to Visuals/Storytelling	High to Data Hygiene/Price/Specs
Interaction Layer	GUI (Website/App)	API / MCP (Data Layer)
Merchandising Key	Visual Adjacency (Layout)	Structured Data & API Latency

Part III: The Friction — Algorithmic Efficiency vs. Human Serendipity

The central conflict identified in this research is the tension between the efficiency of the agent and the serendipity required for brand building and high-margin retail. While agents optimize for relevance, they risk creating a sterile commerce environment that strips away the joy of discovery.

3.1 The Filter Bubble and the Erosion of Serendipity

Critiques of algorithmic curation suggest that over-personalization leads to a “filter bubble” or “information cocoon”. An AI agent tasked with “buying a white shirt” will return the mathematically optimal white shirt based on price, fabric weight, and reviews. It is unlikely to suggest a navy blazer that would look great with it, unless it has been explicitly programmed with a “cross-sell” logic.

This phenomenon has been described as the “commoditization of the search”. By prioritizing relevance and accuracy, the agent eliminates the “noise” of the shopping experience. However, for a retailer, that “noise” is often where the profit lies. The impulse buy—the gum at the register, the scarf next to the coat—is a function of visual interruption. Agents do not have “impulse”; they have directives.

The Serendipity Deficit

Research into recommender systems highlights the difficulty of engineering serendipity—defined as “unexpected yet advantageous findings”. While computational models are being developed to inject “interestingness” and “diversity” into recommendations, current agentic models largely optimize for precision (did I get exactly what I asked for?) rather than delight (did I find something I didn’t know I wanted?).

This poses a severe threat to “window shopping.” If consumers stop browsing and start delegating, the opportunity for brands to communicate “soft attributes” (values, personality, heritage) diminishes. An agent cannot “feel” the prestige of a luxury brand’s heritage; it can only read the “founded_date” field in a database. If the pricing and specs do not align with the agent’s logic, the brand is filtered out, regardless of its cultural cachet.

3.2 Just-in-Time (JIT) Discovery: The New Merchandising Logic

However, it is reductive to view agents solely as destroyers of merchandising. They also offer a new form of merchandising innovation: Just-in-Time (JIT) Discovery.

Traditional JIT principles, pioneered by Toyota in manufacturing, focused on minimizing inventory waste by having parts arrive exactly when needed. Agentic commerce applies this logic to consumption. The agent acts as a supply chain manager for the household.

Predictive Replenishment: Instead of a consumer “stocking up” (pantry stuffing), the agent monitors usage and orders replenishment to arrive precisely before stock-out.
Context-Aware Merchandising: Agents can analyze real-time contexts (weather, local events, travel plans) to surface products “just in time.” For example, an agent noting a flight booking to Aspen might autonomously suggest purchasing thermal wear two weeks before the trip.

For the retailer, this data-driven approach allows for “Just-in-Time” inventory strategies, reducing waste and improving cash flow. The innovation here is not in visual discovery, but in temporal discovery. The agent discovers the need before the human consciously acknowledges it.

3.3 The “Soft Attribute” Problem

A significant hurdle for agentic commerce is the quantification of “soft attributes.” Human shoppers evaluate products based on complex, often irrational criteria: “Is this brand ‘cool’?” “Does this packaging feel premium?” “Does this company align with my political values?”.

Current AI agents struggle to weigh these factors against “hard attributes” like price and delivery speed. While agents can be programmed to filter for “sustainable” or “cruelty-free” products, nuances like “brand coolness” or “aesthetic compatibility” are difficult to encode in a JSON file. This creates a vulnerability for brands that rely on emotional storytelling. If they cannot translate their “vibe” into data that an agent can parse and value, they risk being commoditized.

Part IV: Deep Dive — The Holiday Stress Test

The holiday shopping season serves as the perfect crucible to test the limits of agentic commerce. It is a period characterized by a unique blend of high-stress “deal hunting” and high-emotion “gift selection,” making it an ideal laboratory for observing the human-agent dynamic.

4.1 Shifting Behaviors: The Deal Sniper vs. The Thoughtful Gifter

Traditionally, the holidays are defined by ritual and connection, not just transactions. Consumers rank family time and tradition over gift-giving. However, the 2025 landscape shows a bifurcation in behavior driven by AI adoption.

The Rise of “SantaGPT”

Survey data indicates a massive surge in AI adoption for holiday shopping. In 2025, nearly two-thirds (64%) of Americans expressed willingness to use AI tools or agents to assist with holiday shopping, a dramatic increase from previous years. Specifically, 75% of consumers plan to use AI to find deals, and 67% for gift ideas.

This signals the rise of the “Deal Sniper.” Shoppers are arriving with stronger intent and narrower budgets, utilizing agents to compare options side-by-side and find the absolute lowest price across the entire web. The agent acts as a mercenary, stripping away the festive “fluff” to secure the asset at the lowest cost.

The “Perfect Gift” Paradox

Despite this efficiency, there is deep ambivalence about delegating the selection of gifts. While 63% of consumers trust AI as a “holiday gift guide”, the emotional labor of gifting—the “thoughtfulness” signal—remains a uniquely human domain.

Interestingly, survey data reveals a generational divide: Gen Z consumers are more likely to react negatively (20%) to receiving a gift chosen by AI compared to Baby Boomers (5%). This suggests that for younger cohorts, the effort of discovery is a component of the gift’s value. If an agent picks the gift, the “thoughtfulness” premium evaporates. This validates the “assortment” merchandising principle: the story behind the product matters as much as the product itself.

4.2 The Disruption of Holiday Timelines

Agents are also disrupting the traditional holiday calendar. The “Turkey 12” (the 12 days surrounding Thanksgiving) remains a peak, but shopping is spreading earlier into October and later into December.

Agents facilitate this spread by acting as “always-on” price monitors. A consumer can set an agent to “watch” a specific toy or gadget in October and execute the purchase the moment it hits a target price, whether that happens on Black Friday or a random Tuesday in November. This flattens the traditional “doorbuster” spikes into a continuous, algorithmic search for value, forcing retailers to rethink their promotional cadences.

4.3 Merchandising for the Machine During Holidays

Retailers are responding by optimizing their “digital shelves” for machine readability. During the holidays, “out of stock” is the ultimate sin. Agents will instantly switch to a competitor if a product is unavailable. Therefore, JIT inventory management becomes a critical holiday capability.

Furthermore, retailers are leveraging their own agents to guide holiday shoppers. Walmart, for instance, uses its AI capabilities to predict demand and manage inventory flow during peak periods, reducing stockouts by 30%. This “owned agent” strategy (discussed further in Part VI) attempts to keep the consumer within the retailer’s ecosystem by offering the same efficiency as a third-party agent but with a bias toward the retailer’s own assortment.

Part V: The Economics of Agent-Mediated Retail

The shift to agentic commerce fundamentally alters the economic model of retail, impacting everything from customer acquisition costs (CAC) to inventory management and gross margins.

5.1 The Threat of Disintermediation and Margin Erosion

The most immediate economic threat is disintermediation. As consumers delegate discovery to third-party agents (e.g., OpenAI’s Operator, Google’s Gemini, Amazon’s Rufus), retailers lose the direct traffic that allows for cross-selling and brand building.

If an agent mediates the transaction, the retailer’s website risks becoming merely a fulfillment node—a “dumb pipe” for products—rather than an experience destination. This loss of direct interface separates the brand from the customer data, blinding the retailer to browsing behaviors that inform future product development.

Commoditization and the “Race to the Bottom”

When agents compare products primarily on structured data (price, specs, delivery), brands are forced into a “race to the bottom.” The “brand tax”—the premium a consumer pays for a recognized logo—may evaporate if the agent determines that a generic alternative is functionally identical and 30% cheaper. Agents effectively commoditize categories that were previously differentiated by marketing.

Example: A consumer asks for “high-quality ibuprofen.” An agent, analyzing chemical composition and manufacturing standards, will likely select the generic store brand over the name brand, as the “data” shows they are identical. The name brand’s marketing budget is rendered useless against the agent’s logic.

5.2 The Efficiency Dividend: Inventory, Returns, and Conversion

Conversely, agentic commerce offers significant economic upside in operational efficiency.

Reduced Returns: Product returns are a massive drag on retail profitability, costing billions annually. AI agents that utilize precise sizing data, virtual try-on integration, and “best fit” logic can drastically reduce return rates. By ensuring the product matches the user’s specific needs before purchase, the agent acts as a quality assurance filter.
Inventory Optimization: By enabling JIT purchasing and predicting demand with higher accuracy, agents help retailers avoid overstocking and obsolescence. This reduces working capital requirements and storage costs, directly impacting the bottom line.
Higher Conversion Probability: Traffic driven by AI agents is reported to be significantly higher intent. While direct traffic volume may decrease, the quality of that traffic increases. Salesforce notes that AI-driven discovery traffic converts at a much higher rate because the “browsing” and filtering have already been done by the agent. The “Agent Conversion Rate” becomes a new, critical KPI.

Table 3: Economic Impact of Agentic Commerce on P&L

Metric	Impact Direction	Mechanism
Direct Web Traffic	⬇️ Decrease	Consumers interact with the Agent interface, not the Brand site.
Conversion Rate	⬆️ Increase	Agents pre-qualify products; traffic is high-intent and ready to transact.
Customer Acquisition Cost (CAC)	⬆️ Increase	Competition for “Algorithmic Attention” raises costs; “bidding” for agents.
Return Rates	⬇️ Decrease	Better matching of product specs/sizing to user needs.
Gross Margin	⬇️ Decrease	Price transparency and comparison forces competitive pricing; loss of “brand premium”.
Inventory Carrying Cost	⬇️ Decrease	JIT replenishment and predictive stocking reduce overhead.

Part VI: Strategic Imperatives for Retailers

To survive in an agent-mediated world, retailers cannot simply rely on traditional merchandising. They must adopt a dual strategy: optimizing for the agent (technical readiness) while doubling down on the human (experiential differentiation).

6.1 Optimization for “Algorithmic Attention” (AEO)

Just as retailers spent the last decade optimizing for SEO (Search Engine Optimization), they must now pivot to AEO (Answer Engine Optimization) and Agent Readiness.

Data Hygiene as Strategy

Agents rely on structured data. If a product’s weight, dimensions, material, origin, or compatibility are not explicitly defined in the metadata, the agent cannot “see” it. “Clean, structured, and honest product data” is now a prerequisite for visibility. Retailers must ensure their catalogs are “machine-reasonable,” not just human-readable.

API-First Commerce

Retailers must expose their inventory and pricing via robust APIs to allow agents to query and transact in real-time. This involves adopting standards like the Model Context Protocol (MCP) to ensure seamless integration with third-party agents. The “speed of answer” from an API will become a ranking factor; if an agent has to wait 2 seconds for a stock check, it will move to the next retailer.

Cultivating Algorithmic Trust

Retailers must cultivate “algorithmic trust.” Agents will “punish” retailers who frequently stock out, ship late, or have high return rates by downgrading them in future recommendations. Reliability becomes a marketing asset.

6.2 The “Owned Agent” Strategy

To avoid disintermediation, major retailers are deploying their own “Super Agents” to capture the customer interaction layer.

Case Study: Walmart: Walmart has deployed a suite of agentic AI tools. “Sparky,” their customer-facing agent, assists shoppers with discovery and decision-making, effectively keeping the query within Walmart’s walls. Internally, they use agents (“Marty”) for inventory management, which has reduced out-of-stocks by 30%. By owning the agent, Walmart ensures its assortment is prioritized.
Case Study: Sephora: Sephora utilizes AI for hyper-personalization. Their tools (Virtual Artist, Smart Skin Scan) act as digital beauty consultants. By analyzing past purchases and skin profiles, these agents provide “best fit” recommendations that foster loyalty. Sephora’s “MACH” (Microservices, API-first, Cloud-native, Headless) architecture allows these agents to access real-time inventory and customer data, creating a seamless experience.
Case Study: Nike: Nike employs AI-driven virtual assistants and try-on technology to solve the “sizing” problem—a key barrier in apparel e-commerce. By offering personalized coaching and product matching through their apps, Nike creates a service layer that third-party agents cannot easily replicate, reinforcing direct-to-consumer (DTC) channels.

6.3 Defensive Measures: The “Walled Garden” Paradox

Some brands, particularly in the luxury sector, may choose to opt-out of the agentic ecosystem to preserve exclusivity. By blocking AI crawlers or restricting API access, brands attempt to force consumers to visit their owned channels.

This “Walled Garden” approach is a high-stakes gamble. It preserves the “brand experience” and the high-touch visual merchandising that justifies premium pricing. For a brand like Hermes or Chanel, where the purchase is the experience, this makes sense. However, for mass-market or “masstige” brands, invisibility to agents equates to invisibility to the consumer. If a user asks their agent for “best running shoes” and Nike blocks the agent, Nike simply ceases to exist in that user’s consideration set.

6.4 New KPIs for the Agentic Era

Retailers must adopt new Key Performance Indicators (KPIs) to measure success in this environment.

Agent Conversion Rate: The success rate of agent-driven interactions. How effectively does the brand convert an agent’s query into a sale?
Uniqueness Index: A measure of how distinct the catalog is. High uniqueness protects against commoditization by agents.
API Latency / Uptime: Technical metrics that now directly correlate with sales visibility.
Return on Ad Spend (ROAS) -> Return on Algorithm (ROA): Measuring the efficiency of investments in data structuring and API optimization.

Conclusion: The Hybrid Future

The dichotomy between “deal shopping” and “assortment discovery” is not being erased by agents; it is being codified and amplified. AI agents are becoming the ultimate “deal hunters,” ruthlessly optimizing for efficiency, price, and logistics on commodity goods. They are the new “procurement officers” for the household, stripping away the marketing veneer from toilet paper, batteries, and basic apparel.

In response, human behavior will likely bifurcate. Consumers will happily delegate the drudgery of commodity purchasing to agents (the “deal” mode), but they may retreat further into “assortment discovery” for categories where emotion, taste, and identity matter—fashion, gifts, home decor.

The Verdict on Innovation

Does agentic shopping innovate on merchandising or just optimize conversion? The answer is both.

It innovates by introducing Just-in-Time Discovery, a temporal form of merchandising that finds the need before the human does.
It optimizes conversion by acting as a high-intent filter, reducing returns and waste.

However, it currently fails to replicate the serendipity of the human “hunt.” For the retailer, the path forward is not to choose between the human and the machine, but to build for both. They must build machine-readable catalogs for the agents that replenish the pantry, while simultaneously building human-delighting experiences (both digital and physical) for the moments when shopping is not a chore, but a joy.

The retailers that fail to bifurcate their strategy—attempting to sell commodities with storytelling or luxury goods with bare-bones data—will find themselves invisible to the agent and irrelevant to the human. In 2025, the shelf is digital, the shopper is a bot, but the ultimate consumer remains, stubbornly and importantly, human.