Will Agentic Commerce Give Rise to Direct to Consumer?

For two decades, e-commerce aggregators built multi-billion-dollar tollbooths on top of consumer attention. They did not win by manufacturing superior products, nor did they win by operating cheaper logistics networks. They won because searching thousands of individual brand websites requires cognitive effort that human consumers refuse to expend. By aggregating product discovery and guaranteeing consumer trust under a single search bar, platforms like Amazon became the default gateway for digital trade.

The rise of autonomous buying agents dismantles this advantage. An AI agent does not suffer from cognitive fatigue, does not scroll through sponsored search results, and does not rely on marketplace star ratings to evaluate supplier legitimacy. When a buying agent can inspect inventory, verify technical specifications, negotiate pricing, and execute payments directly across thousands of independent supplier endpoints in milliseconds, the central marketplace loses its reason to exist. Intermediaries survive by owning discovery and trust. An agent that can evaluate a supplier directly needs neither, which is a problem for everyone standing in the middle.

Intermediaries extract rent by monopolizing discovery and trust

The intermediary layer operates as a tax on the distribution of physical goods. Retail marketplaces charge merchants not for the cost of hosting data, but for access to human attention. In 2025, Amazon recorded $172.2 billion in third-party seller services revenue out of $716.9 billion in total net sales, according to Ecom Brainly’s Amazon Seller Statistics. When combining referral fees, fulfillment charges, inbound placement fees, and mandatory pay-per-click advertising required to stay visible in search results, platforms extract between 40% and 50%+ of a third-party seller’s gross merchandise value, as detailed by Nova Analytics.

Brands submitted to this extraction because the alternative was obscurity. Building a direct-to-consumer (DTC) channel required spending equivalent sums on digital ad networks to attract human web traffic. Aggregators unbundled the merchant from the end customer, turning brand owners into white-label suppliers competing for placement on a platform-owned interface.

This model relies on a structural limitation: human buyers cannot query the open web efficiently. Humans require centralized search indexes, standardized product pages, and consolidated checkout carts. Intermediaries built businesses by converting that human inefficiency into a platform fee. What search engines did to media discovery, retail marketplaces did to commerce distribution. When the interface shifts from a human staring at a web browser to an autonomous agent executing code, the economic justification for a 40% platform take-rate disappears.

Buying agents evaluate specifications instantly but collapse on experience

The shift toward agent-driven purchasing is accelerating across consumer and business channels. Generative AI traffic to US retail sites surged 4,700% year-over-year from July 2024 to July 2025, as recorded in Chargeflow’s Agentic Commerce Report. According to McKinsey & Company, AI agents are projected to orchestrate up to $1 trillion in US B2C retail revenue and $3 trillion to $5 trillion globally by 2030. Commercetools research highlights Bain & Company estimates placing US agentic commerce at $300 billion to $500 billion by 2030, while Gartner analysis estimates that 20% of all digital commerce transactions will be executed through AI platforms or buying agents by 2030.

Buying agents excel at evaluating structured product parameters. An agent can query thousands of merchant databases simultaneously, parse technical data sheets, compare unit prices across shipping zones, and verify inventory availability without human intervention. The mechanics of programmatically querying product systems are standardized through open technical standards such as the Model Context Protocol Architecture. For standardized and objective purchases, agentic evaluation renders marketplace aggregation redundant.

Agents struggle when forced to evaluate subjective human experience. Standardized specifications fail to capture aesthetic preference, tactile feel, brand alignment, or nuanced customer dispute resolution. The limits of autonomous agent operations were demonstrated by Klarna’s customer service deployment. As documented in Klarna’s OpenAI Partnership Announcement, their AI assistant handled 2.3 million conversations in its first month—equivalent to two-thirds of customer inquiries—and reduced average resolution times from 11 minutes to under 2 minutes. However, an analysis by the AI Professionals Directory notes that Klarna subsequently walked back its AI-only framing and re-hired human support agents after customer satisfaction degraded on complex, high-stakes disputes.

AI agents excel at deterministic catalog search and factual verification. They fail when executing open-ended subjective judgments.

Cryptographic identity and payment protocols replace central marketplace arbitration

Critics argue that direct-to-consumer trade cannot scale without centralized marketplaces because consumers rely on platforms to manage financial risk, handle chargebacks, and aggregate physical shipping. This view misunderstands how payment infrastructure and logistics networks are unbundling from discovery aggregators.

Financial liability in automated commerce is shifting from platform arbitration to cryptographic verification. According to Flagright’s Chargeback Economics Analysis, global chargeback losses are projected to reach $33.79 billion in 2025 and $41.69 billion by 2028, with dispute volume rising from 261 million to 324 million. Traditional fraud prevention relies on human clickstream signals—such as browser fingerprints, mouse movements, and IP geolocation—that do not exist during agentic purchases. As noted by Chargeflow’s Merchant Liability Guide, merchants historically absorbed this financial risk when automated interactions lacked human authorization proof.

To solve this gap, open protocol architectures are establishing machine-level identity and liability handshakes. Google, Shopify, and payment processing networks introduced cryptographic spending mandates through the Universal Commerce Protocol Specification and the Agent Payments Protocol (AP2). Similarly, Stripe and OpenAI co-developed the open-source Agentic Commerce Protocol Specification, allowing agents to complete purchases inside conversational workflows. Direct-to-consumer brands such as Glossier and Spanx integrated this standard via Shared Payment Tokens (SPTs) to process instant checkouts without routing buyers to traditional web pages, as covered by KuCoin’s Stripe Agentic Suite Release.

Autonomous verification stacks now bypass central marketplace trust mechanics entirely. Architectures detailed by Nevermined’s Skyfire Review show how “Know Your Agent” (KYA) identity tokens and dedicated agent wallets combine with universal execution APIs like Rye. As announced in Skyfire’s Universal Checkout Launch, buying agents can prove identity, pass anti-bot defenses, verify merchant inventory, and transfer funds directly to independent DTC storefronts.

Logistics aggregation is undergoing a parallel unbundling. Modern third-party logistics (3PL) networks allow independent brands to offer decentralized fulfillment, competitive shipping rates, and two-day delivery windows without surrendering inventory to marketplace warehouses. Buying agents can aggregate multi-merchant orders programmatically at the software layer, optimizing shipping routes across distinct DTC APIs. Software coordinate logistics across independent networks, eliminating the need for a central distributor to manage physical fulfillment.

The market splits between machine-queried commodities and brand-led experiences

The shift toward agentic commerce will not transform every product category uniformly. The commercial landscape is bifurcating into machine-queried commodity purchases and high-touch brand experiences.

Commodity goods—including household goods, standardized replacement parts, consumer electronics accessories, and routine nutritional supplements—will shift heavily toward direct machine purchasing. For these categories, product discovery is an engineering calculation based on material composition, delivery speed, seller reputation scores, and total cost. When a consumer delegates replenishment of these items to a buying agent, the marketplace search bar becomes irrelevant. The agent queries brand endpoints directly, compares terms, and executes payment across direct channels. Standardized commodity brands that maintain proprietary direct-ordering APIs will bypass marketplace commissions, passing a portion of those savings to the customer while capturing higher operating margins.

High-touch experience goods—including luxury fashion, cosmetics, artisanal crafts, and tailored apparel—follow a different trajectory. These categories rely on sensory evaluation, physical fit, personal identity, and emotional storytelling. An agent cannot feel the drape of a fabric or judge how a fragrance develops over time. While an agent can manage the operational mechanics of sizing checks, returns processing, and price tracking, human buyers will retain creative control over initial discovery.

What cloud infrastructure did to enterprise software, buying agents are doing to retail distribution channels. High-touch brands will use DTC channels to build direct emotional relationships with human buyers, while optimizing backend APIs so those buyers’ agents can execute frictionless repeat transactions.

Storefronts must transition from visual layouts to structured machine endpoints

To remain competitive in an Agentic Internet, direct-to-consumer brands must redesign their digital infrastructure. The traditional e-commerce storefront—optimized for human eyeballs with visual banners, promotional pop-ups, and conversion funnels—is invisible to a software buying agent.

Brands must treat their direct storefronts as machine-readable APIs. The primary requirements for agentic brand legibility include:

1. Granular Machine-Readable Schema Data: Implementing structured data across all product pages following the Schema.org Product Type Specification. Storefronts must expose explicit attributes including gtin, exact stock availability (offers), currency-denominated pricing (priceCurrency), and standardized return policies (hasMerchantReturnPolicy).
2. Open Model Context Protocols: Exposing real-time inventory databases, shipping calculators, and variant structures via the Model Context Protocol Architecture. This enables LLM-driven agents to inspect stock levels without scraping web interfaces.
3. Native Agentic Checkout Protocol Support: Integrating standardized checkout workflows through the Agentic Commerce Protocol Specification and Universal Commerce Protocol Standards. Systems must accept cryptographically signed mandates and Shared Payment Tokens (SPTs) to support instant agent-driven settlements.
4. Agent Engine Optimization (AEO): Transitioning search strategy from traditional SEO to machine optimization. Storefronts must supply deterministic factual data to generative search systems, ensuring buying agents evaluate the brand accurately during automated supplier selections.

We do not merely build prettier storefronts for human consumers; we engineer the underlying growth operations and protocol integrations that make brands legible to machine buyers.

The digital storefront used to be a visual catalog for human browsers. The next one is an API for buying agents.