Every technological shift in commerce has restructured the value chain. The internet moved power from physical retailers to platforms. Mobile moved power from desktops to app stores. Social commerce moved power from search engines to social networks. Each shift created new control points - positions in the value chain where an actor could extract disproportionate value by controlling information flow, customer access, or transaction infrastructure. Agentic commerce is the next shift, and it is creating the most consequential new control point in a generation: the agent layer.
This essay examines how the traditional commerce value chain is being restructured by autonomous agents. It maps the new control points, analyses the dynamics of platform lock-in through trust ownership, confronts the unresolved merchant-of-record question, and argues that the value chain of the agentic era will be governed not by attention or distribution, but by trust architecture.
I. The Traditional Commerce Value Chain and Its Assumptions
The traditional commerce value chain is a linear sequence: manufacturer produces goods, distributor aggregates and moves them, retailer presents them to customers, and the customer selects and purchases. Digital commerce compressed this chain - manufacturers could sell direct, platforms could aggregate millions of sellers, and customers could compare across the entire market from a single interface - but the fundamental assumption remained: the customer is a human being who discovers, evaluates, and purchases through a visual interface.
This assumption shaped every layer of the value chain. Marketing exists to capture human attention. Visual merchandising exists to influence human perception. Checkout flows exist to reduce human friction. Customer support exists to resolve human confusion. Loyalty programmes exist to exploit human psychology. The entire infrastructure of digital commerce is optimised for a human customer who sees, feels, hesitates, and decides.
The platforms that dominate digital commerce - Amazon, Shopify, Google Shopping, Meta Commerce - captured their positions by controlling the chokepoints in this human-centric value chain. Amazon controls the marketplace where humans discover products. Google controls the search engine where humans begin their purchase journey. Meta controls the social graph where humans are influenced. Shopify controls the infrastructure through which merchants serve humans. Each platform’s power derives from its position between the human customer and the merchant.
When the customer is an agent, these chokepoints shift. The agent does not discover products through Amazon’s marketplace - it queries structured data feeds. It does not begin its journey on Google - it evaluates merchants against its principal’s mandate. It is not influenced by social media - it processes trust signals. The platforms that control the human-centric value chain do not automatically control the agent-centric value chain. The control points must be re-established - and the contest to establish them is already underway.
II. The Agent Layer as a New Control Point
The agent layer is the new chokepoint in the commerce value chain. It sits between the human principal’s intent and the merchant’s fulfilment. The agent receives the human’s mandate, evaluates available merchants, selects the best option, and executes the transaction. Every piece of information that flows between the human and the merchant passes through the agent. Every transaction that connects the human’s intent to the merchant’s product passes through the agent. The agent is the gatekeeper.
The economic power of a gatekeeper depends on the alternatives available to the actors on either side. If the human can easily switch agents, the agent’s power is limited. If the merchant can easily reach humans without the agent, the agent’s power is limited. But if the human’s trust is concentrated in a single agent platform - if the human has configured their preferences, trained their agent, and accumulated a transaction history that would be costly to replicate elsewhere - the switching cost is high. And if the merchant cannot reach agents without the agent platform’s intermediation, the merchant is dependent.
This is the classic platform economics dynamic, applied to the agent layer. The agent platform that accumulates the most human trust, the most transaction history, and the most merchant integrations becomes the dominant intermediary. It can extract value through transaction fees, preferred placement, data monetisation, and subscription revenue. The agent layer becomes the new Amazon - not a marketplace of products, but a marketplace of trust.
The candidates for this position are already visible. OpenAI’s ChatGPT is building agent capabilities that mediate between humans and services. Google’s Gemini is integrating agent functionality into the search and commerce ecosystem. Apple’s Siri is positioned as the agent layer for the Apple ecosystem. Amazon’s Alexa is evolving from a voice assistant to an autonomous purchasing agent. Each of these platforms is competing to become the trust intermediary between humans and commerce - and the winner will control the most valuable chokepoint in the agentic value chain.
III. Platform Lock-In Through Trust Ownership
Platform lock-in in traditional digital commerce operates through network effects, data accumulation, and switching costs. Amazon locks in customers through Prime membership, purchase history, and the convenience of a single marketplace. Google locks in advertisers through search data, audience targeting, and the dominance of the search channel. These lock-in mechanisms are well understood.
Agentic commerce introduces a new and more powerful form of lock-in: trust ownership. When a human delegates purchasing authority to an agent, they are placing trust in the agent’s judgment, reliability, and alignment with their interests. This trust is not transferable. It is built through experience - through successful transactions, accurate recommendations, and reliable execution. The human’s trust in their agent is a form of relationship capital that accumulates over time and cannot be easily replicated by a competing agent platform.
Trust ownership creates lock-in that is deeper than data lock-in or network effect lock-in. A human can export their purchase history from one platform to another. They cannot export their trust. They cannot transfer the accumulated evidence that their agent understands their preferences, respects their constraints, and acts in their interests. Switching agents means starting the trust-building process from scratch - and in a world where the agent makes consequential financial decisions on the human’s behalf, the cost of trusting a new, unproven agent is high.
From the AXD perspective, trust ownership lock-in raises fundamental design questions. Delegation design must address whether trust data should be portable - whether a human should be able to transfer their trust history from one agent platform to another. Trust architecture must address whether trust should be platform-specific or protocol-level - whether trust credentials should be issued by platforms or by independent trust authorities. These are not just design questions. They are questions about the economic structure of the agentic era.
IV. The Merchant-of-Record Question
When a human purchases a product from a merchant, the legal and financial relationships are clear. The merchant is the seller. The human is the buyer. Consumer protection law, warranty obligations, and dispute resolution mechanisms all assume this direct relationship. The merchant-of-record - the entity legally responsible for the transaction - is the merchant.
When an agent purchases on behalf of a human, these relationships become ambiguous. The agent selected the merchant. The agent evaluated the product. The agent executed the transaction. The human may not have been directly involved in any of these decisions - they delegated the authority and the agent acted within its mandate. If the product is defective, who is responsible? If the agent selected a merchant that misrepresented its product data, is the agent platform liable? If the human’s mandate was ambiguous and the agent interpreted it in a way the human did not intend, who bears the cost?
The merchant-of-record question is not merely a legal technicality. It determines the economic incentives of every actor in the agentic value chain. If the merchant bears full liability for agent-mediated transactions, merchants will demand more control over the agent’s evaluation process - more structured data requirements, more verification steps, more explicit consent from the human principal. If the agent platform bears liability, the platform will impose stricter standards on merchants and extract higher fees to cover the liability risk. If the human bears liability for their agent’s actions, the human will demand more transparency and control over the agent’s decision-making.
The AXD perspective is that the merchant-of-record question must be resolved through delegation design. The delegation - the act by which the human grants authority to the agent - must specify the scope of authority, the constraints on the agent’s decisions, and the allocation of liability. A well-designed delegation makes the merchant-of-record question answerable: the liability follows the authority. If the agent acted within its delegated authority, the human bears the commercial risk. If the agent exceeded its authority, the agent platform bears the risk. If the merchant misrepresented its data, the merchant bears the risk. The delegation is the contract that governs the value chain.
V. Disintermediation and Re-Intermediation
Every shift in commerce technology produces a cycle of disintermediation and re-intermediation. The internet disintermediated physical retailers but re-intermediated through platforms like Amazon. Mobile disintermediated desktop commerce but re-intermediated through app stores. Each cycle removes one set of intermediaries and replaces them with another.
Agentic commerce is producing the same cycle. The agent layer disintermediates traditional platforms - the human no longer needs Amazon’s marketplace to discover products, because the agent evaluates merchants directly. But the agent layer re-intermediates through agent platforms - the human now depends on the agent platform to mediate their relationship with commerce. The intermediary has changed, but intermediation persists.
The question is whether this re-intermediation is inevitable, or whether the design of the agentic value chain can produce a more distributed structure. Open protocols - like the Agent Commerce Protocol, the Universal Commerce Protocol, and the Machine Payments Protocol - represent an attempt to create a protocol-level trust infrastructure that does not require platform intermediation. If trust credentials are issued by independent authorities, if product data is structured according to open standards, and if payment protocols are interoperable, then agents can transact with merchants directly - without a platform controlling the chokepoint.
The outcome depends on whether trust becomes a platform asset or a protocol asset. If trust is platform-specific - if an agent’s trust in a merchant is meaningful only within the context of a particular agent platform - then platform re-intermediation is inevitable. If trust is protocol-level - if trust credentials are portable, verifiable, and interoperable across platforms - then the value chain can be more distributed. The design of trust architecture will determine the economic structure of agentic commerce.
VI. The Trust-Governed Value Chain
The value chain of the agentic era will be governed by trust. Not by attention, which governed the advertising-funded web. Not by distribution, which governed the platform era. Not by data, which governed the personalisation era. By trust - the structural foundation on which every transaction, every relationship, and every delegation is built.
In the trust-governed value chain, the actors that invest in trust infrastructure capture disproportionate value. Merchants that build machine-readable trust credentials, transparent policies, and reliable fulfilment capture direct relationships with agents. Agent platforms that build trustworthy, transparent, and aligned agent systems capture human trust. Protocol designers that build interoperable, verifiable, and portable trust standards capture the infrastructure layer. The value chain rewards trust investment at every layer.
The AXD Institute’s position is that the trust-governed value chain is not just an economic prediction. It is a design imperative. The discipline of trust architecture exists precisely to design the trust relationships that govern the agentic value chain. The discipline of delegation design exists to design the authority flows that determine liability, value capture, and economic power. The economic structure of agentic commerce is downstream of design decisions. The designers of trust will be the architects of the new economy.