Retail Drift: Who Owns the Customer When When AI Shops for Them

Platforms are partnering with AI shopping agents while investing in owned portals and fulfillment networks.

Nov 06, 2026

Across retail and digital commerce, platform executives have spent the past year preparing for agentic shopping, where autonomous artificial intelligence assistants compare prices, assemble baskets, and initiate checkouts on behalf of consumers. Yet late-October and early-November 2026 earnings calls highlighted a strategic tension taking shape beneath that vision.

While marketplaces and commerce providers are actively partnering with third-party AI assistants to capture new demand, analysts are increasingly questioning what happens to direct customer relationships and high-margin retail media advertising if discovery moves off-platform.

Rather than betting on a single outcome, major platforms are hedging: DoorDash announced a pending $17 billion agreement to acquire Instacart to expand its grocery reach and retailer services, Shopify acquired the shop.com domain while opening its merchant infrastructure to external AI agents, and Apple and Visa outlined plans to anchor agentic commerce around operating-system assistants and verified payment credentials.

In today’s Retail Innovation Week newsletter, we examine why agentic disintermediation is emerging as a key strategic question on Wall Street, how platforms are balancing open AI partnerships against owned destinations, and how physical retailers are using rapid store-based fulfillment and tactile flagship spaces to strengthen direct customer loyalty.

This newsletter has been developed using insights from PSFK’s Fodda, an expert-led context layer available for use by agencies, brands and retailers in their research and AI workflows. Ask Fodda to create a newsletter just like this for you, today.

Who Controls the Customer When AI Shops?

As AI agents begin searching for products, comparing prices, and completing purchases on behalf of consumers, commerce platforms face a fundamental question: who controls the customer relationship when shoppers no longer need to visit their apps or websites? For marketplaces that depend on direct traffic, high-margin retail media advertising, and repeat engagement, the shift could challenge established business models. Yet it also creates opportunities for companies that control the underlying technology, payment systems, and physical fulfillment networks behind each transaction.

Recent earnings calls reveal how major commerce businesses are preparing for this possibility, even as actual agent-driven purchasing remains at an early stage.

DoorDash’s proposed $17 billion acquisition of Instacart, announced on October 29, 2026, would bring together two major grocery and local delivery networks. DoorDash CEO Tony Xu presented the pending transaction primarily as an opportunity to improve grocery fulfillment, expand merchant services, and compete more effectively with larger conglomerates. Yet the Q3 2026 call conversation quickly turned to what happens when third-party AI assistants begin directing orders: pressed by Cantor Fitzgerald analyst Deepak Mathivanan on the risk of agentic disintermediation, and by Rothschild & Co Redburn analyst Dominic Ball on how external AI assistants are affecting order flow, Xu acknowledged the long-term potential while disclosing that third-party consumer agents had generated only several hundred DoorDash orders in total so far. Rather than treating external assistants purely as competitors, Xu noted that DoorDash is exploring win-win partnerships and agent-based advertising alongside its own shopping tools.

Shopify is approaching the same shift from another direction, and its November 3, 2026 third-quarter call exposed two distinct tensions around consumer discovery and AI unit economics. Questioned by RBC Capital Markets analyst Paul Treiber on the strategic rationale behind Shopify’s roughly $100 million purchase of the shop.com domain, President Harley Finkelstein framed the move as an expansion of its centralized consumer shopping destination, even as the company enables independent merchants to sell directly through third-party AI agents via protocols like WebMCP and shop.dev. Meanwhile, when MoffettNathanson’s Michael Morton pressed CFO Jeff Hoffmeister on whether scaling the Sidekick merchant AI assistant would pressure Subscription Solutions gross margins, Hoffmeister reaffirmed Shopify’s long-standing 80 percent gross margin philosophy and revealed that Sidekick’s compute cost per conversation fell 28 percent quarter-on-quarter (and 18 percent lower in total during Q3 2026), suggesting that AI unit economics are improving even as JPMorgan’s Bryan Smilek probed management’s Q4 infrastructure investments for 2027 merchant adoption.

The question of who controls an AI-mediated purchase extends well beyond retail marketplaces. Apple is developing Siri as an operating-system-level interface for AI assistance to anchor discovery above individual apps, while Visa is positioning authenticated payment credentials and transaction security as essential to autonomous purchasing. In Southeast Asia, Grab emphasizes the protective value of its integrated merchant, driver, and hyper-local delivery network.

Taken together, these developments suggest that agentic commerce may not eliminate intermediaries so much as change which intermediaries matter. Consumer-facing apps could lose some influence over discovery, while the platforms providing trusted payments, merchant catalogs, and reliable physical fulfillment become increasingly essential.

What Happens When Discovery Moves to AI Assistants?

The strategic debate surfacing across recent earnings calls centers on how digital marketplaces capture high-margin revenue beyond basic fulfillment and transaction fees. Over the past five years, platforms including DoorDash, Instacart, Grab, and Amazon built powerful profit engines on retail media advertising, charging brands and merchants for sponsored search slots and promotional placements inside their apps.

In PSFK’s assessment, widespread adoption of third-party AI shopping agents introduces a structural challenge to that model: if consumers delegate routine search and basket assembly to external assistants, platforms risk losing the high-margin browsing sessions that fuel retail media, unless they adapt their advertising models for agentic interfaces or retain direct consumer touchpoints.

Behind this platform anxiety lies a fundamental transformation in how products are discovered and evaluated. When an AI assistant handles product search, traditional visual merchandising, banner ads, and keyword bidding lose much of their power. In its report Who Wins When AI Does the Shopping? New Evidence on Agentic Commerce, McKinsey argues that winning customer consideration increasingly requires AI-Ready Product Information: structured, machine-readable product attributes, real-time inventory feeds, and verified customer feedback that an algorithm can parse directly. Brands that rely on eye-catching packaging or paid digital shelf space could find themselves invisible to an autonomous buyer if their data cannot be indexed by the machine.

At the same time, this shift toward algorithmic assistance is altering the nature of consumer loyalty. Shoppers are increasingly comfortable delegating routine, functional replenishment to automated filters, yet that convenience does not automatically translate into emotional brand connection. In its 2027 Global Consumer Predictions, Mintel examines these Authority Shifts, pointing to a growing divide in how consumers assign trust: routine purchases are handed off to algorithmic efficiency, while destination brands must work harder to demonstrate authentic craft, human perspective, and physical presence. As Mintel’s complementary 2027 Global Wellness Predictions observes under The Recommendation Engine Is the New Brand, when an algorithm curates daily choices, the recommendation tool itself becomes a powerful proxy for brand authority, forcing consumer brands to find new ways to stand out beyond functional utility.

Crucially, however, automating product discovery does not solve the physical challenges of commerce. Even when an AI agent successfully identifies the optimal product at the lowest price, the transaction still depends on physical delivery. In its e-commerce study The New Formula for Black Friday, DHL highlights how delivery-related friction remains a primary cause of cart abandonment at checkout, even as shoppers eagerly use AI tools to hunt for deals. If delivery windows are uncertain, shipping fees are prohibitive, or local inventory is inaccurate, the transaction breaks down regardless of how sophisticated the front-end agent may be.

For retail and platform leaders, the implication is clear: autonomous shopping agents may reshape the digital discovery layer, but durable commercial power remains anchored to trusted product data, distinctive brand equity, and dependable physical fulfillment networks that can execute where software stops.

Open Partnerships, Owned Portals & Infrastructure Rails

Rather than treating autonomous agents purely as adversaries, digital commerce leaders are pursuing dual-track strategies that combine open AI integrations with investments in owned portals, physical logistics, and payment rails:

In local delivery and grocery aggregation, DoorDash, Instacart, and Grab are pairing external AI partnerships with deeper retail software and fulfillment scale. Before agreeing on October 29, 2026 to be acquired by DoorDash in the pending $17 billion transaction, Instacart demonstrated that commerce aggregators can collaborate directly with frontier AI providers by integrating with Anthropic’s Claude to help users turn meal plans into shoppable grocery carts. At the same time, DoorDash CEO Tony Xu explained during the Q3 2026 call that the pending Instacart combination is designed to give retailers of every size the technology, merchant software, and fulfillment density needed to compete against larger tech conglomerates. When Citizens analyst Andrew Boone and Wolfe Research analyst Shweta Khajuria asked how the pending deal could strengthen fulfillment over time, Xu pointed to the long-term potential of pairing Instacart’s grocery relationships and Storefront Pro software (which powered Aldi USA’s March 2026 e-commerce redesign) with DoorDash’s physical DashMarts locations. In Europe, DoorDash’s Wolt subsidiary has similarly partnered with Tesco in Prague on Tesco Hned, a rapid grocery delivery pilot operating across 28 stores with target delivery times around 40 minutes. In Southeast Asia, Grab CEO Anthony Tan emphasized during its Q3 2026 call that multi-segment integration drives resilience: “With all three core segments now profitable and our outlook raised for the second time this year, Grab has more engines contributing to profit than at any point in our history.”

In merchant e-commerce infrastructure, Shopify is simultaneously opening its merchant catalog to third-party AI agents while strengthening its own consumer-facing ecosystem. During Shopify’s Q3 2026 earnings call on November 3, 2026, President Harley Finkelstein described Shopify as adept at operating in two realities at once: supporting how merchants sell today while building for where commerce is heading next. On one track, Shopify is actively enabling third-party agentic shopping through developer protocols such as WebMCP, shop.dev, and integrations with AI platforms including Muse, Instinct, and Claude Duet. On the second track, Shopify spent roughly $100 million to acquire the shop.com domain and continues to scale its Shop app and Shop Pay network (now reaching 400 million consumers). In PSFK’s view, this dual approach allows Shopify to process transactions whether a shopper buys through an external AI assistant or browses directly inside Shopify’s owned consumer portal, while improving Sidekick’s unit compute efficiency (down 28 percent per conversation quarter-on-quarter) protects Subscription Solutions margins.

In consumer hardware and payment networks, Apple and Visa are building foundational layers intended to sit underneath agentic transactions regardless of which application initiates the purchase. During Apple’s Q4 FY2026 earnings call on November 2, 2026, CEO John Ternus highlighted the role of system-level intelligence: “We see tremendous opportunity in AI as well. Siri AI marks a new chapter in AI at Apple. It is a completely reimagined assistant that is powerful, personal, private, and deeply relevant to our users’ daily lives.” Alongside Apple’s operating-system assistant, Visa CEO Ryan McInerney emphasized during Visa’s October 27, 2026 call that security and verified identity will anchor autonomous commerce: “We’ve seen meaningful growth acceleration this year, and as I look ahead to 2027, one thing is clear, trust is the next growth vector for Visa.” While neither company frames these initiatives solely as defensive measures, PSFK notes that controlling OS-level intent and tokenized payment credentials gives both companies durable leverage as agentic shopping matures.

Defensive Measures: Store-Based Fulfillment Speed and Tactile Flagships

While digital platforms balance third-party agent partnerships against owned portals, physical retailers are responding along two distinct tracks: using store proximity to accelerate last-mile fulfillment, and redesigning flagship spaces around customization and hospitality that reward in-person visits. PSFK has been tracking how retailers across grocery, home improvement, beauty, luxury, and athletic apparel are executing these two strategies:

On the fulfillment front, major grocers and big-box retailers are pairing conversational AI discovery with store-based delivery networks. As an early mover in March 2026, French grocer Carrefour integrated a conversational shopping assistant into ChatGPT in France to help customers plan meals and assemble baskets, while routing final checkout back to Carrefour’s own website. In North America, big-box retailers are using their physical store footprints to deliver rapid fulfillment: PSFK’s research has highlighted how Walmart expanded its 30-minutes-or-less Express Delivery service to 38 U.S. markets with physical stores fulfilling 80 percent of all e-commerce orders, while Home Depot rolled out 3-hours-or-less express delivery across more than 2,000 U.S. stores. Dense local inventory close to the consumer remains an operational asset whether an order begins on a retailer’s app or through an external AI assistant.

On the experiential retail front, beauty, luxury, and lifestyle brands are investing in physical customization and hospitality that give shoppers a reason to visit stores directly. In Seoul’s Myeongdong district, PSFK has observed how Amorepacific-owned skincare brand Laneige opened its renovated experiential flagship store in 2026 following an 18-month overhaul, featuring three in-person customization services: a bespoke Lip Sleeping Mask bar, the Neo custom shade-matching cushion service, and the Cream Skin lab for tailored toners. While custom formulation does not eliminate digital competition, it shifts the flagship visit from commodity restocking to personalized service. Similarly, Spanish luxury house LOEWE opened a two-floor, 344-square-meter fragrance flagship in Seoul complete with scent galleries and an upper-floor café, Lululemon reopened its Cheongdam flagship in Seoul with an integrated movement studio for community workouts, and New Balance debuted its flagship at The Grove in Los Angeles accompanied by a temporary listening lounge and café activation at launch.

Taken together, these two retail tracks underscore a practical division of labor: store-based micro-fulfillment captures convenience-driven replenishment, while experiential flagships cultivate brand affinity through consultation, customization, and community spaces.

In a modern AI-generated interpretation applying eighteenth-century political economy to digital commerce, foundational economist Adam Smith offers a classical lens on market intermediaries and the division of labor. Drawing on The Wealth of Nations, an economist applying Smith’s framework today would argue that when an intermediary collects rent primarily by standing at a transactional bottleneck, any technology that lowers the cost of connecting buyers and producers challenges that margin. When his Classic Agent is queried through Fodda, this contemporary AI interpretation suggests that digital intermediaries naturally respond by investing in physical warehousing, last-mile fulfillment, and settlement rails that pure software routing cannot easily replicate. Viewed through that classical lens, durable commercial advantage flows from productive fulfillment capital and specialized service rather than a purely digital tollgate.

Examining this shift from a modern brand-building perspective, go-to-market strategist and Banquet Chief Executive Officer Jess Graham diagnoses the rush toward autonomous shopping through the lens of the Seamlessness Paradox and Texture. The former head of global consumer product marketing at Instagram would argue that equating algorithmic retrieval with genuine human discovery is a category error: as AI shopping agents optimize strictly against explicit queries, they compress evaluation into a backend call and push digital discovery toward an undifferentiated algorithmic mean. When her Human Agent is queried through Fodda, it explains that full transaction automation puts undifferentiated brands on a measurable Brand Decay Rate, making intentional Texture (the friction worth keeping, where attention slows and serendipity happens) essential. As agents absorb upper-funnel replenishment, post-purchase touchpoints, unboxing, community spaces, and physical retail rooms become the primary arenas where lasting brand loyalty is built.

2027 Playbook for Retail & Platform Leaders

As agentic shopping evolves from early pilot volumes into a mainstream channel, retail and platform executives can balance open AI discovery with direct customer ownership across four priorities:

Balance Open Agent Integrations With First-Party Checkout Incentives: Rather than blocking external AI assistants or relying exclusively on owned portals, test conversational discovery partnerships (as Instacart and Shopify have done) while using loyalty rewards, personalized pricing, and seamless fulfillment guarantees (as Carrefour piloted in March 2026) to encourage final checkout and account linkage on first-party channels. Explore AI checkout and retail partnership strategies.
Track AI Compute Unit Economics Alongside Conversion Lift: Following Shopify’s Q3 2026 disclosure that Sidekick’s cost per conversation dropped 28 percent quarter-on-quarter while lifting early merchant order milestones by 3 percent, benchmark generative and agentic AI features on both unit compute efficiency and measurable merchant or basket conversion gains. Analyze e-commerce AI margin and efficiency metrics.
Pair Digital Discovery With Store-Based Express Fulfillment: Use existing store networks to deliver sub-hour and same-day fulfillment (mirroring Walmart’s 38-market 30-minute delivery footprint and Home Depot’s 2,000-plus store express network), ensuring that whether an order starts in a native app or an external AI assistant, local inventory density wins the conversion. Study store-based express fulfillment networks.
Invest in Tactile Flagship Services and Post-Purchase Texture: Complement automated replenishment by dedicating physical retail footprints to in-person customization (like Laneige in Myeongdong), hospitality (like LOEWE Seoul), and movement spaces (like Lululemon Cheongdam) that deepen emotional brand equity beyond the search bar. Explore experiential flagship store formats.

Advertising Drift: Agencies’ Blind Bet on AI Helps Clients Squeeze Fees Fast

More Fodda insight in our PSFK Weekly newsetter: What happens when an industry invests heavily in technology that teaches its customers to expect its services for less? Over the past two years, global advertising holding companies including Omnicom Group, Publicis Groupe, Stagwell, and WPP enthusiastically embraced artificial intelligence, staging flashy demonstrations of automated copywriting, instant visual production, and programmatic media optimization.

But corporate procurement departments were watching closely: If an algorithm can generate forty campaign variants in three seconds, enterprise brand executives reason, there is no longer any commercial justification for funding monthly retainers that bill for human hours.

In the PSFK Weekly newsletter, we examine an industry trapped in a crisis of its own making. As holding companies race to rebrand as enterprise software providers, they are colliding with skeptical clients, conflicts of interest over data sovereignty, and an ironic operational reality: artificial intelligence has not eliminated agency labor, but it has convinced clients that the work should be free.

Full Story

The 30 Seconds Connection Plan

I’d love to start a conversation with anyone involved in Connections Planning. This aspect of the advertising process not my forte but users are asking the Fodda tool for specific marketing recommendations so I am exploring how we could generate such plans based on a trend or an insight.

Questions users can ask Fodda today to get such a plan: “Using Fodda, generate a full Audience-Led Brand Connections Plan for On Running around Gen Z social running clubs in the US”

Questions I’m asking: Is this process helpful? Useful? What would be additive? What speed of delivery feels too fast? What workflows could we integrate into.

Email me at piers@psfk.com - Piers, PSFK Founder & Fodda Founda

Query Fodda

Explore the corporate disclosures, PSFK research insights, and Human Agent frameworks cited in this issue directly in the Fodda Context Layer:

Discussion about this post

Ready for more?