The Agent Shopper: Brand Strategy for the Agentic Commerce Era

1. The Agent Shopper Has Arrived

The pace of change is striking. In December 2025, Klarna launched its Agentic Product Protocol—an open standard that made over 100 million products across twelve markets instantly discoverable by AI agents. Weeks later, Google announced the Universal Commerce Protocol (UCP—an open standard for agent-merchant interoperability) at the National Retail Federation conference, backed by more than twenty partners including Shopify, Walmart, Target, Mastercard, and Visa. OpenAI and Stripe co-developed the Agentic Commerce Protocol (ACP), an open standard for AI-enabled payments and commerce—though OpenAI’s initial attempt at direct checkout inside ChatGPT was quietly discontinued in early March 2026 after low transaction completion rates, a revealing signal that consumer purchasing behavior in agent environments is still forming. Perplexity expanded its AI shopping assistant to more than 5,000 merchants through PayPal integration. And in a revealing legal confrontation, a federal court in California granted Amazon a preliminary injunction against Perplexity in March 2026, finding that AI agents accessing Amazon’s systems on behalf of users did not constitute authorized access by the retailer’s standards.

These are not incremental upgrades. They represent the emergence of a new commercial architecture in which a growing share of buying decisions are evaluated, filtered, and sometimes executed by systems operating on behalf of human buyers—what this article calls the agent shopper. The scale projections are substantial: Gartner projects that 40% of enterprise applications will incorporate task-specific AI agents by the end of 2026, up from less than 5% in 2025. Adobe reported a 693% increase in traffic to U.S. retail sites from generative AI tools during the 2025 holiday season, with AI-referred shoppers converting at 31% higher rates and delivering 254% higher revenue per visit. Morgan Stanley estimates that agentic commerce could reach $190–385 billion in U.S. e-commerce spending by 2030. McKinsey frames the global opportunity at $3–5 trillion in AI-orchestrated commerce revenue by the same year. For brand leaders, these figures raise a question that decades of marketing theory did not prepare them for: what happens to brand strategy when the buyer is not a person but an agent shopper acting on a person’s behalf?

2. Who Builds the Agent Shoppers—and What Constitutes Their Intelligence

Before asking what brands should do, executives need to understand who is building the agent shoppers—because most of them are not being built by brands. Third parties define the agent’s intelligence; brands must respond.

The Builders

Five categories of firms are constructing the agent ecosystem. AI platforms and assistants (OpenAI, Google, Apple) control conversational interfaces and sit at the moment of intent capture. Google’s UCP connects Gemini and AI Mode in Search directly to merchant backends. These platforms are building the general-purpose agent that interprets what the user wants and orchestrates the path to fulfillment. Commerce infrastructure providers (Stripe, payment networks, identity services) build the transaction plumbing beneath the agent’s decisions—how purchases are authorized, routed, and completed. Retailers, marketplaces, and merchant ecosystems are building their own agent-compatible systems. Klarna’s protocol makes merchant catalogs machine-readable. Walmart deploys its own Sparky assistant while simultaneously integrating with Google’s Gemini and OpenAI’s ChatGPT—a deliberate strategy to secure decision slots across multiple agent ecosystems. Aggregators and data providers organize supply, comparison, and review data at scale, serving as normalized, trusted sources of commerce data for agent shoppers. And brands themselves—but selectively. The pattern of digital history suggests caution: consumers have consistently migrated to general-purpose tools that serve their interests. But some brands will build narrower agents around high-consideration categories, concierge service, replenishment, or post-purchase guidance.

What Constitutes Agent Intelligence

An AI shopping agent is not a blank search engine. It carries a set of core components that together constitute what we call agent intelligence—the full set of design parameters, functional capabilities, and learning inputs that determine how an agent shopper decides. Understanding these components is critical because the question of who influences each determines who holds power in agentic commerce.

The Brand Intelligence framework (Sun, 2026) provides the conceptual parallel. The Command Center—the brand’s algorithmic decision engine—generates real-time marketing decisions by optimizing who to engage, what to offer, when to act, where to reach them, and how to present it, all to maximize User Lifetime Value (ULTV) over the entire customer relationship. The consumer’s agent shopper solves a symmetric problem: who to buy from, what to buy, when to buy, where to transact, and how to execute—optimizing for whatever objective the user defines. The seven components fall into three layers—Architecture (the agent’s design parameters), Capabilities (its functional reach), and Learning (the inputs that make it smarter over time).

Together, these seven components constitute agent intelligence—the commercial mind of the agent shopper. The three-layer classification matters strategically: Architecture determines which brands can compete for the agent’s attention. Capabilities determine which brands the agent can actually evaluate and transact with. Learning determines which brands the agent develops loyalty to over time.

3. The Five Decisions an Agent Shopper Makes — and What Brands Must Feed Them

Before exploring how agent decisions reshape brand strategy, it helps to anchor the analysis in what the agent shopper is actually deciding. The agent makes five sequential decisions at the moment of commerce, each demanding a different form of machine legibility from the brand.

1. Discovery: “Does this brand belong in my consideration set?”

A binary gate — inclusion or exclusion — based on category relevance, attribute match, and minimum data completeness. If the brand doesn’t pass this gate, nothing else matters. The legibility requirement is findability: standardized taxonomy, structured attributes, presence in agent-facing feeds, and Generative Engine Optimization (GEO) readiness — the nine tactics for making a brand’s knowledge base optimized for AI retrieval and generation (Sun, 2026, Ch. 10).

This is where decision-slot competition occurs — the struggle to be included in the agent’s shortlist before the consumer ever sees the options. In traditional digital commerce, brands competed on the search page or in the feed. In agentic commerce, the competition happens inside the agent’s reasoning process, and the losers are never seen by the consumer.

2. Evaluation: “How does this rank against alternatives?”

The agent scores and compares across specifications, price, reviews, service terms, and reputation signals. This is where the objective function bites — the same alternatives produce different rankings depending on whether the agent optimizes for cost, quality, sustainability, or minimal regret. The legibility requirement is comparability: standardized attributes in formats agents can score, transparent pricing that includes total cost of ownership, and machine-readable differentiation.

Here objective-function governance becomes a strategic issue. The brand cannot control which platform the consumer uses or how that platform’s default logic interprets an ambiguous instruction like “find me the best value.” But it can ensure that its advantages are legible to whichever interpretation prevails — price, satisfaction, risk minimization, or sustainability. The brands whose differentiation is only asserted in advertising copy will lose to brands whose differentiation is evidenced in structured, verifiable data.

3. Configuration: “What specific option best fits this user?”

The agent selects the variant, bundle, add-ons, service level, and delivery option that match the user’s constraints and preferences. This goes beyond ranking to constructing the actual offer. The legibility requirement is modularity: product data structured so the agent can explore combinations, compatibility rules, personalization parameters, and bundle logic.

This is where composability competition emerges. The brand that offers rigid, fixed SKUs loses to the brand whose offer architecture is modular enough for the agent to construct the optimal combination for a specific user. The competition is not about having the best single product — it is about exposing the richest solution space. A brand with ten configurable parameters gives the agent more degrees of freedom to find a fit than a competitor with a take-it-or-leave-it listing. In agentic commerce, the most composable offer wins the configuration decision.

4. Transaction: “At what terms, when, and how?”

The agent decides timing, channel, payment method, and — in more advanced interactions — negotiated terms (price, warranty, bundle trade-offs, delivery commitments). The legibility requirement begins as actionability (real-time inventory, delivery windows, payment protocol compatibility) and evolves toward negotiability: programmable deal parameters, price floors and ceilings, concession logic, escalation triggers.

This is the arena of deal-space competition. Brands no longer compete solely on the price they set — they compete on the parameter space within which their systems can negotiate. One brand publishes a fixed price and waits. Another exposes programmable deal logic: price floors and ceilings, bundle trade-offs, loyalty incentives, delivery flexibility, warranty upgrades. The agent will gravitate toward the richer negotiation space because it can optimize across more dimensions simultaneously.

5. Trust calibration: “How much do I trust this brand, and how do I update that trust?”

This cuts across every other decision. The agent assigns a confidence weight to each brand based on accumulated signals and updates that weight after every interaction. The legibility requirement is verifiability: observable data the agent can incorporate into its confidence model. This is what builds agent-trust equity over time — and it compounds.

Unlike the other four decisions, trust calibration creates confidence-model competition — a form of rivalry that accumulates rather than resets. Every fulfilled promise strengthens the brand’s position in the agent’s confidence model; every broken promise weakens it. Unlike human trust, which is fuzzy and forgetful, agent trust is cumulative and precise. The brand with a longer, cleaner performance record earns a structural advantage that new entrants or inconsistent performers cannot quickly overcome. This makes operational excellence not just an efficiency metric but a compounding competitive asset.

Table 1. The Five Agent Decisions, Legibility Requirements, and Competitive Arenas

Agent Decision	Legibility Requirement	Competitive Arena	Primary Tier
Discovery — Does this brand belong in my consideration set?	Findability: standardized taxonomy, structured attributes, GEO readiness, protocol presence	Decision-slot competition: invisible exclusion — losers are never seen	Tier 1
Evaluation — How does this rank against alternatives?	Comparability: standardized formats, transparent pricing, machine-readable differentiation	Objective-function governance: whose definition of “best” controls the ranking	Tier 1–2
Configuration — What specific option best fits this user?	Modularity: modular product data, configurable offers, agent-explorable combinations	Composability competition: richest solution space wins the assembly	Tier 2
Transaction — At what terms, when, and how?	Actionability → Negotiability: programmable deal logic, concession rules, escalation triggers	Deal-space competition: widest negotiation bandwidth wins the terms	Tier 2–3
Trust calibration — How much do I trust this brand?	Verifiability: observable outcome data that feeds the agent’s confidence model	Confidence-model competition: cumulative performance creates compounding advantage	All tiers

How Agent Decisions Reshape the User Journey

The five agent decisions map to, but do not replicate, the five stages of the traditional User Journey: need recognition, information search, evaluation, purchase, and post-purchase experience. Three structural transformations change how brands must compete:

• Time horizon collapses. What used to take days or weeks — the consumer evaluating, comparing, and waiting — now happens in seconds inside the agent. Brands must serve the agent at machine speed.
• Stages run in parallel, not sequence. The agent evaluates, configures, and prices in parallel; human attention is no longer the bottleneck. Brand messaging optimized for sequential funnel stages misses the parallel-processing reality.
• Memory persists. Where humans forgot last quarter’s disappointing experience, the agent remembers exactly. Trust calibration is now a long-term competitive asset, not a marketing impression.

4. The System That Serves Both — If You Built It

Brands now face two customers: the human user and the AI agent acting on their behalf. The strategic question is not whether to build for one versus the other — it is whether your existing intelligence architecture can serve both.

The Brand Intelligence framework (Sun, 2026) describes the architecture that already does. Eight modules — mobile app, online store, smart products, physical stores, online community, customer service, social media, and retail media — feed a Command Center that synthesizes intelligence across the brand-owned ecosystem. The same Command Center that personalizes the human experience can structure the data that agents consume. The architecture does not need to be rebuilt; it needs to be made machine-legible.

Machine legibility is the degree to which a brand’s value is structured, accessible, and interpretable by AI agents. It has four dimensions:

• Structured — product attributes, pricing, inventory, and service terms in schemas the agent can parse
• Accessible — available through agent-compatible protocols (UCP, ACP, MCP) and real-time APIs
• Interpretable — semantic relationships that let the agent understand differentiation, not just specifications
• Verifiable — backed by observable outcomes (delivery, performance, returns) that feed the agent’s confidence model

Brands that have invested in the eight-module architecture have most of this already. The work is not building new infrastructure; it is exposing existing infrastructure in agent-readable form.

This is the Put-and-Take Method (Sun, 2026, Ch. 10) applied to agents. Put: expose structured summaries, comparable specifications, and verifiable claims in formats agents can consume. Take: route agent-referred traffic back into the brand-owned ecosystem where the deeper experience, community, and behavioral data create the next cycle of intelligence. The agent layer becomes a discovery channel for the owned ecosystem — not a replacement for it.

The brands that built the eight-module architecture for human users are now discovering that they built the agent infrastructure too. The brands that delayed will find that catching up on machine legibility requires the same architectural work, only now under competitive pressure.

5. How Agentic Commerce Changes Brand Strategy

Five strategic shifts follow from the agent-shopper architecture. Each runs counter to instincts shaped by twenty years of digital marketing.

From persuasion to verification

Brand marketing has long centered on persuasion — the art of moving humans through emotional appeal, narrative, and brand association. Agents are not persuaded. Agents are persuaded only by what they can verify. Verifiable claims (third-party certifications, independent reviews, structured performance data, transparent supply chain attributes) carry more weight than the most beautifully produced campaign. The marketing budget shifts from emotional reach to evidence infrastructure.

From decision-stage funnel to decision-slot competition

The funnel assumed a human moving through awareness, consideration, decision. Agents don’t move through stages; they execute all five decisions in parallel inside their reasoning process. Competition for brand visibility is no longer at the top of the funnel (awareness) — it is for inclusion in the agent’s shortlist (the decision slot). Brands that aren’t in the agent’s consideration set don’t lose; they don’t exist.

From owned-channel personalization to objective-function governance

For a decade, brands invested in personalizing the experience on their owned channels. That work still matters — but the new frontier is governing the objective function the agent applies. The brand cannot dictate what the agent optimizes for, but it can influence which dimensions of value are visible. A brand whose advantages are evident on price will lose to a brand whose advantages are evident on whatever the agent is optimizing for.

From SKU strategy to composability strategy

The shift from rigid SKUs to modular offer architecture is no longer a manufacturing question; it’s a brand-strategy question. In agentic commerce, the most composable offer wins. Brand strategy now includes architecting the parameter space within which the agent can assemble the right combination for each user — not building the perfect product, but building the perfect range of possibilities.

From advertising spend to operational excellence as a marketing investment

The most surprising shift: operational excellence becomes the marketing budget. Agent-trust equity is built by fulfilled promises, accurate delivery, predictable performance, and consistent service quality. The brand that ships on time builds compounding trust capital in the agent’s confidence model. The brand that promises more than it delivers loses position with every disappointment. Operations and marketing converge.

6. What Executives Should Do Now

The agent-shopper transition is not a five-year horizon. The infrastructure is shipping now, the standards are being set this year, and the brands that are absent from the agent’s reasoning will face invisible exclusion — not visible decline, but quiet absence from the consideration set. Five moves matter most.

1. Audit machine legibility today

For each of your top product categories, conduct an honest assessment: can an agent with no prior knowledge of your brand find, compare, configure, transact, and verify your offer? Score yourself on each of the four legibility dimensions: structured, accessible, interpretable, verifiable. The gap between today’s state and what agents will require is usually larger than executives assume.

2. Adopt agent protocols deliberately

UCP, ACP, MCP, and platform-specific APIs are not engineering decisions — they are strategic decisions about which agent ecosystems your brand will be visible in. Walmart’s deployment across Sparky, Gemini, and ChatGPT is the model: deliberate multi-ecosystem presence rather than betting on a single platform.

3. Restructure product data as a strategic asset

Most brands manage product data as an operational chore. In agentic commerce, structured product data — the catalog, the attributes, the relationships, the verifiable claims — is the brand’s primary surface for the agent. Treat it with the seriousness once reserved for advertising creative.

4. Build programmable deal logic

Move from fixed pricing toward a deal space: parameters the agent can negotiate within. This requires technical infrastructure (pricing engines that expose floors, ceilings, and trade-off logic) but more importantly, the organizational comfort with handing real-time pricing authority to systems rather than humans.

5. Measure agent-trust equity

Add agent visibility, citation rate, and confidence-model metrics to the strategic dashboard. Existing brand-health measures (awareness, consideration, preference) are about humans; they don’t capture agent dynamics. Brands that can measure their position in the agent’s confidence model can defend and grow it. Brands that cannot, can’t.

7. Forward Look

Three developments will determine the pace at which agentic commerce reshapes whole industries.

First, the convergence of agent protocols. Whether UCP, ACP, and MCP converge into a single dominant standard or remain fragmented will determine how much infrastructure each brand must maintain. The early signs suggest convergence is happening faster than past digital-protocol races (think of HTTP vs. SMTP) because the underlying economics — brands wanting visibility, agents wanting interoperability — align.

Second, the maturation of consumer trust in agents. The OpenAI direct-checkout discontinuation is a useful data point: consumers are not yet ready to hand full transactional authority to agents. The middle ground — agents that recommend, configure, and prepare transactions for human approval — is where the next eighteen months will be spent. The brands that build smoothly into this hybrid pattern will define the consumer expectation.

Third, the regulatory frontier. The Amazon-Perplexity court ruling is a preview of what’s coming. Who is allowed to access a brand’s commerce surface, under what authorization, with what data rights, will be litigated across every major market in the next two years. Brands need a legal and protocol position now, not after the precedents are set.

The Brand Intelligence framework was designed for this transition. The same eight modules, the same Command Center, the same Flywheel of Intelligence — reframed for agent legibility instead of human-only engagement. The architecture compounds across both audiences. The brands that have invested in their intelligence architecture are better positioned than they realize. The brands that have not are running out of time.

Cross-References

• Chapter 2: Platform Enterprises — the platform economics underpinning agentic commerce
• Chapter 3: Consumer Decision-Making in the AI Era
• Chapter 4: Building Brand Intelligence — the eight modules and Command Center
• Chapter 7: Smart Products and Smart Stores — the data infrastructure beneath agent-readable commerce
• Chapter 8: Mobile Applications — the brand-owned surface that survives the agent layer
• Chapter 9: The Command Center and Intelligent Marketing — Agentic Marketing, Machine Thinking
• Chapter 10: Omni-Domain Marketing and Brandnetics™ — the GEO playbook applied to agent visibility
• BI-AR-01: Your Brand Is Invisible to AI: Why GEO Is the New SEO
• BI-CS-02: KFC China — intelligent activation across an agent-ready super-app
• BI-CS-04: Walmart — enterprise intelligence and multi-ecosystem agent deployment

References

The Brand Intelligence framework, Agent Shopper, Machine Legibility, Agent Intelligence, Decision-Slot Competition, Objective-Function Governance, Composability Competition, Deal-Space Competition, Confidence-Model Competition, Agent-Trust Equity, Programmable Deal Logic, Put-and-Take Method, Flywheel of Intelligence, and related concepts presented in this work are the intellectual property of Baohong Sun, as published in Brand Intelligence: Navigating the Transformation in the AI and Web3 Era (Springer Nature, 2026). Any use, citation, or adaptation of these frameworks requires proper attribution. License: CC BY-NC-ND 4.0.

Frequently Asked Questions

Quick answers to the questions most readers ask about this piece.

What is agentic commerce?

Commerce conducted by AI agents acting on behalf of consumers — agents that search, compare, configure, transact, and learn over time. The brand-customer funnel becomes a brand-agent funnel.

How does agentic commerce change brand strategy?

Five strategic shifts: from storytelling to structured intelligence, from one brand equity to three (consumer + machine + agent-trust), from asserted to evidenced differentiation, from static pricing to programmable deal logic, and from platform dependence to Put-and-Take sovereignty.

What is the Agent Intelligence Stack?

A three-layer architecture: Agent Learning (preference inputs, outcome data), Agent Capabilities (product graph, transaction protocols), and Agent Architecture (objective function, decision logic, guardrails). Brands have the most direct influence on Learning.

What is the Put-and-Take Method in agentic commerce?

A method where brands "put" structured, agent-readable data into agent ecosystems for reach, then "take" agent-referred traffic back into brand-owned ecosystems for data sovereignty. The brand gains reach without ceding the customer relationship.

What should CMOs do first to prepare for agentic commerce?

Five actions, in order: audit machine legibility today, adopt agent protocols deliberately (MCP, A2A), restructure product data as a strategic asset, build programmable deal logic, and measure agent-trust equity as a third brand-equity dimension.