Chapter 19: Conversational Commerce and Shopping Assistants |
Summary |
Conversational commerce has moved far beyond the scripted chatbots of the previous decade. Today's shopping assistants are capable of multi-turn reasoning, personalized recommendations, and in many cases, autonomous transaction execution. Amazon's Rufus and Alexa for Shopping now serve hundreds of millions of customers, combining product intelligence with the ability to build carts, set price thresholds, and complete purchases automatically. Walmart has deployed its own agentic framework, including the Sparky assistant and partnerships with OpenAI and Google that embed Walmart's catalog directly into and Gemini. Meanwhile, beauty retailers like Sephora, Fenty, and Ulta are racing to claim the AI shopping moment through conversational interfaces on WhatsApp, , and Google's AI surfaces. This chapter examines the architectures, capabilities, and strategic implications of these systems across multiple industries, and explores what the shift toward agentic commerce means for both retailers and consumers. |
|
1. The Three-Stage Maturity Curve |
To understand where conversational commerce stands today, it helps to trace how it evolved through three distinct stages. |
Stage One: Scripted Chatbots. The first generation of conversational tools in retail was built for support deflection. A customer would type a question like 'where is my order,' and a rule-based bot would match keywords to a canned response. These systems were reactive, memoryless, and brittle. Ask a follow-up question and the bot would reset to zero. They handled the queries their designers anticipated and failed on everything else . |
Stage Two: Context-Aware Conversational AI. The second stage introduced memory and reasoning. Modern large language models allowed assistants to carry context across a session, understand intent rather than keywords, and generate responses dynamically. A shopper could ask about fit, compare alternatives, or check availability without hitting a dead end. This is the stage most major retailers have now reached, though with varying degrees of sophistication . |
Stage Three: Agentic Commerce. The third stage, now emerging, is where the assistant does not just discuss a purchase but executes it. An agentic shopping assistant can add items to a cart, apply promotions, complete checkout, and even purchase on the customer's behalf when predefined conditions are met. This is a meaningfully different capability from a recommendation engine that links out to a product page. It requires integration with inventory systems, payment infrastructure, and identity verification . |
The distinction matters because vendors often use 'conversational AI' to describe both Stage Two and Stage Three capabilities. The practical difference is autonomy: can the system take action, or does it merely inform |
|
2. Amazon: Rufus, Alexa for Shopping, and the Scale of Conversational Retail |
Amazon operates the largest conversational commerce deployment in the world. Its AI shopping assistant, originally launched as Rufus, has been used by more than 250 million customers, with monthly average users up 149 percent and interactions up 210 percent over the past year. Customers who use the assistant while shopping are over 60 percent more likely to make a purchase during that session . |
In May 2026, Amazon merged Rufus with Alexa Plus to create Alexa for Shopping, which now occupies the primary search bar on Amazon.com and in the mobile app. The integration represents a significant escalation: instead of a separate chat window, the assistant is embedded in the most prominent real estate Amazon has . |
2.1 Technical Architecture |
Alexa for Shopping is built on Amazon Bedrock, drawing on a mix of large language models including Anthropic's Claude Sonnet, Amazon Nova, and a custom model trained on Amazon's product catalog, customer reviews, community Q&As, and web data. A real-time router selects the appropriate model for each query type, optimizing for capability, latency, and answer quality. The system also uses Retrieval-Augmented Generation to pull insights from sources like The New York Times, USA Today, and Vogue when answering questions about products and trends . |
2.2 Key Capabilities |
The assistant's capabilities span the entire shopping journey. On the discovery end, it can answer broad questions like 'what do I need for a hiking trip' and generate a shoppable checklist. For product evaluation, it produces side-by-side comparisons drawing on specifications, reviews, and expert sources. For purchase execution, it can reorder past items, build carts, and set price alerts . |
The most advanced feature is automatic purchasing based on customer-set thresholds. A user can instruct the assistant: 'Add this sunscreen to my cart if the price drops to $10 and I haven't purchased it in the last two months.' The system then monitors the price and executes the purchase when conditions are met, without further human intervention . |
The assistant also has account memory, meaning it retains information across sessions. If a customer has previously shared that they have young children who play sports or a dog that sheds, the assistant factors those details into recommendations. Ask about holiday gifts for kids, and it will suggest age-appropriate sports books and apparel. Ask about a robot vacuum, and it will highlight pet hair cleanup as a key feature . |
2.3 Cross-Device Continuity |
Because Alexa for Shopping is integrated with Alexa Plus, context carries across devices. A conversation started on an Echo smart speaker about a science project can continue on Amazon.com when the customer types 'show me what I need to buy for my science project.' Price alerts set via voice appear in the app and on Echo Show displays. Amazon describes this as 'cross-device continuity' and it represents a meaningful step toward a unified shopping identity that persists regardless of interface . |
|
3. Walmart: Sparky, Super Agents, and Platform Partnerships |
Walmart has taken a different but equally aggressive approach to conversational commerce, building internal agentic infrastructure while simultaneously distributing its catalog through third-party AI platforms. |
3.1 Sparky: The In-App Shopping Assistant |
Walmart's customer-facing AI assistant is called Sparky. Integrated into the Walmart app, Sparky helps shoppers find and compare products, build lists, get personalized recommendations, and plan for occasions by synthesizing reviews and product data into clear responses. Behind the scenes, multi-agent orchestration and fallback handling enable Sparky to support customers from discovery through checkout . |
Walmart has framed Sparky as part of a broader 'agentic commerce' strategy. The company recently introduced a framework of four internal 'super agents' designed to serve customers, associates, partners, and developers. The customer-facing agent is Sparky; the others handle tasks ranging from customer support routing to supply chain coordination . |
A notable feature of Sparky's design is its emphasis on explainability. Rather than simply recommending a product, it explains why. A recommendation might read: 'Recommended because this cereal has the lowest sugar content, is organic certified, and is on promotion today.' Walmart measures Sparky's success not by conversational metrics but by business KPIs: search conversion rate, click-through rate, cart addition rate, and net promoter score . |
3.2 and Gemini Integrations |
Walmart has partnered with both OpenAI and Google to make its catalog shoppable directly within those companies' AI assistants. |
The OpenAI partnership, announced in October 2025, allows users to browse and purchase Walmart products directly on using a 'buy' button. The catalog includes apparel, entertainment, packaged food, and other products from Walmart and Sam's Club . |
The Google partnership, announced in January 2026, embeds Walmart's assortment into Gemini. Customers can build a basket and purchase directly within the Gemini app, with Walmart handling order fulfillment. The integration supports account linking so that recommendations can be based on past Walmart purchases, and items purchased via Gemini can be combined with existing Walmart or Sam's Club carts . |
Walmart's incoming CEO, John Furner, framed the strategy plainly: 'The transition from traditional web or app search to agent-led commerce represents the next great evolution in retail.' The company's goal, he said, is to 'close the gap between I want it and I have it' . |
|
4. Beauty and Fashion: The Race to Own the AI Shopping Moment |
The beauty and fashion sectors have been among the most aggressive adopters of conversational commerce, partly because their products benefit enormously from the kind of personalized guidance that AI assistants can provide. |
4.1 Sephora Inside |
Sephora launched an app inside , piloting in the United States. Customers can ask for beauty advice in the chat and receive curated recommendations based on preferences from their Beauty Insider account. Sephora's global community of more than 80 million active members provides the data foundation that makes personalization viable. Payment and checkout within the app are planned for future updates . |
4.2 Fenty Beauty on WhatsApp |
Fenty Beauty built an AI-powered beauty adviser called 'Rose Amber' for WhatsApp, marking its first formal partnership with the platform in the U.S. Users can ask questions about skin concerns, take quizzes, and explore products across Fenty Beauty, Fenty Skin, and Fenty Hair. The assistant responds with product suggestions alongside creator videos and customer reviews . |
The WhatsApp channel carries particular weight outside the United States. In Brazil, more than 20 percent of L'Oreal's direct-to-consumer online sales flow through conversational commerce on WhatsApp, and the company has found that the platform converts abandoned carts at six times the rate of email . |
4.3 Ulta Beauty and Google's AI Surfaces |
Ulta Beauty has taken a platform-agnostic approach, making its assortment shoppable across Google's AI surfaces. The retailer is rolling out agentic commerce within AI Mode in Search and the Gemini app, where shoppers can receive recommendations, compare options, and complete streamlined checkout directly within Google's conversational interfaces. The integration uses Google's Universal Commerce Protocol, an open standard for agentic commerce. On its own properties, Ulta is launching Ulta AI, a shopping assistant built on Gemini Enterprise that draws on insights from its 46 million-plus members . |
4.4 Luxury and the Brand Identity Challenge |
For luxury brands, the shift to conversational commerce raises unique challenges around brand control. Swap, an agentic commerce platform, has developed 'agentic storefronts' that mirror a brand's existing site while adding interactive AI agents for search, styling, virtual try-on, and checkout. For Paul Smith, Swap captured the voice of the designer himself and turned it into an agent that customers can converse with about designs and styling . |
The tension is between reach and control. As Swap's CEO observed, 'everyday essentials will be bought through an agent or . But for those products where you have a connection to the brand, you're still going to want that brand experience.' Brands that can provide an elevated conversational experience on their own terms may retain more customer relationships than those that cede the interface entirely to third-party platforms . |
|
5. The Infrastructure of Agentic Commerce |
Behind the consumer-facing assistants lies a less visible but equally important layer of infrastructure. Agentic commerce requires protocols and standards that allow AI agents to access product catalogs, check inventory, process payments, and verify identity across organizational boundaries. |
5.1 Protocols and Standards |
Several competing and complementary standards are emerging. Google's Universal Commerce Protocol, used by Ulta Beauty, provides an open standard for agentic transactions. OpenAI and Stripe have developed their own protocols for in-chat purchases. Shopify has deployed 'agentic storefronts' that allow millions of merchants to appear natively within and other AI surfaces without custom integration work . |
The fragmentation is a concern. J.P. Morgan Payments has noted that the industry will need to align on standard protocols that enable agents to access merchant product catalogues, inventory, pricing, and post-sale data. Open questions include the definition of consumer consent in autonomous transactions, liability when an agent misinterprets a prompt, and data use standards across consumers, merchants, and platforms . |
5.2 Payments and Trust |
Payment providers are positioning themselves as neutral infrastructure for agentic commerce. J.P. Morgan Payments has stated that its focus is to create solutions that support merchants whether they host their own agents or distribute products through third-party consumer agents. The goal is flexibility across an evolving landscape . |
Consumer trust remains a gating factor. Research indicates limited willingness to delegate transactions due to concerns around security, privacy, and reliability. There is a gap between the influence AI assistants have on purchase decisions and their authority to execute those decisions without human approval . |
|
6. Beyond Retail: Conversational Commerce in Other Industries |
While retail and e-commerce have led adoption, the patterns established there are spreading to adjacent industries. |
6.1 Travel and Hospitality |
Conversational AI is being deployed for trip planning, booking, and itinerary management. The same agentic capabilities that allow a shopping assistant to build a cart and check out can be applied to assembling flight, hotel, and activity packages. The challenge is greater because inventory is more fragmented and pricing more dynamic. |
6.2 Financial Services |
Banks and financial institutions are exploring conversational interfaces for account management, product recommendations, and even transaction execution. The stakes are higher than in retail, and regulatory requirements around consent and disclosure are more stringent. But the same agentic architecture applies: understanding intent, accessing account data, and executing authorized actions. |
6.3 Healthcare and Wellness |
Pharmacies and health platforms are using conversational AI for medication management, appointment scheduling, and wellness product recommendations. The personalization capabilities demonstrated in beauty retail translate directly: skin concerns, product sensitivities, and routine preferences can all inform recommendations. |
6.4 Telecommunications |
Telecom providers are deploying conversational assistants for plan selection, device recommendations, and account management. The product comparison capabilities that Amazon and Walmart have built for physical goods apply to service plans with different data allowances, international options, and device bundles. |
The common thread across industries is the shift from scripted support to agentic action. Any domain that involves researching options, comparing alternatives, and executing a transaction is a candidate for conversational commerce. |
|
7. The Strategic Implications for Retailers |
The rise of agentic shopping assistants reshapes competitive dynamics in several ways. |
Discovery is shifting. For two decades, the starting point for online shopping was a search engine or a retailer's own site. Now, consumers are increasingly asking AI assistants for product recommendations. If an agent does not surface a retailer's product, that product is not ranked lower; it is absent . |
Data quality becomes decisive. Agentic commerce systems evaluate products through structured data: specifications, availability, pricing, delivery times. Retailers that invest in clean, standardized product data will be more visible to agents. Rich imagery and persuasive copywriting, while still valuable for human shoppers, matter less when the decision-maker is an algorithm . |
Control of the interface determines who owns the customer. Retailers face a strategic choice: build their own conversational experiences and retain customer relationships, or distribute through third-party platforms and sacrifice some control for reach. Walmart has chosen both paths simultaneously. Amazon, with its integrated Rufus-Alexa assistant, has largely kept the interface in-house . |
The ROI calculation changes. Traditional AI projects are often justified by cost reduction. Agentic commerce projects are better evaluated by revenue metrics. Walmart's Sparky team measures search conversion, comparison conversion, and return rates. Amazon measures purchase likelihood among users of the assistant. The question is not 'how much did we save' but 'how many more customers bought, and how much more did they spend' . |
|
8. Detailed Summary |
Conversational commerce has matured from keyword-matching chatbots to agentic systems capable of autonomous action. This chapter examined the architectures, deployments, and strategic implications of this shift across multiple industries. |
Amazon operates the largest deployment through Rufus and its successor, Alexa for Shopping. Built on Amazon Bedrock with a mix of LLMs, the assistant serves hundreds of millions of customers, offering personalized recommendations, product comparisons, account memory, and automatic purchasing based on customer-set thresholds. Its integration into the primary Amazon search bar represents the most prominent positioning of conversational AI in retail . |
Walmart has built Sparky as an in-app assistant emphasizing explainable recommendations and journey-based design, while simultaneously distributing its catalog through and Gemini. The company's four-super-agent framework represents an unusually systematic approach to agentic infrastructure, spanning customer, associate, partner, and developer roles . |
Beauty and fashion retailers have been aggressive adopters. Sephora launched inside , Fenty built a WhatsApp adviser, and Ulta connected its catalog and loyalty data to Google's AI surfaces. Luxury brands are exploring agentic storefronts that preserve brand identity while adding conversational commerce capabilities . |
Infrastructure for agentic commerce is emerging through protocols from Google, OpenAI, Stripe, and Shopify, though fragmentation remains a concern. Payment providers are positioning as neutral enablers. Consumer trust and willingness to delegate transactions are still gating factors for full autonomy . |
Beyond retail, the same agentic patterns are spreading to travel, financial services, healthcare, and telecommunications. Any domain involving research, comparison, and transaction is a candidate. |
Strategic implications include the shift of discovery to AI interfaces, the growing importance of structured product data over persuasive content, the choice between owned and distributed conversational experiences, and a shift in ROI measurement from cost savings to revenue outcomes. |
The trajectory is clear: conversational commerce is becoming the primary interface for an increasing share of digital transactions, and the retailers and brands that build effective agentic capabilities will be positioned to capture that shift. |
cs@easiersoft.com