Agentic Commerce.
When your customer isn't human.

Sometime in late 2025, a structural shift happened to commerce that almost nobody priced in. Not the launch of a new AI model, not a regulatory change, not a viral product. The shift was quieter and more permanent. The agent layer arrived, and customers started letting it shop on their behalf. Not as a novelty. Not as an experiment. As a default. The implications for brand strategy are larger than the implications of e-commerce itself, and most categories are still building for a buyer that is no longer the one making the decision.

The headline number from our tracking is the one most operators will fixate on, so let us address it first. Autonomous purchasing agents now mediate a measurable share of household transactions across our 2,400-merchant monitoring panel. The trajectory points to roughly thirty-one percent of qualifying purchases passing through some form of agentic intermediary by the third quarter of this year. The number is large. The number is real. The number is not the story.

The story is what the agent is doing differently from the human it replaces, and what that means for every brand that built its positioning on the assumption that the decision-maker would respond to emotion, identity, design, or aesthetic loyalty. The agent does not. The agent responds to a different set of signals entirely. And the brands that are quietly outperforming in this transition are the ones that figured out which signals matter, often without naming it as agent strategy at all.

What the agent actually does.

The mental model most brand teams hold of agentic commerce is wrong in a useful way. The model assumes the agent is a smarter shopper. A better-informed version of the human, making the same kind of decision the human would, just faster. That framing is comforting because it suggests the existing brand toolkit still applies, only sharpened. It does not.

The agent is not a smarter shopper. The agent is a fundamentally different category of buyer. Its evaluation function looks more like a procurement department than a consumer. It weighs structured data heavily. It treats unverified claims as zero-weighted. It cross-references your stated specifications against third-party sources before it accepts them. It does not respond to scarcity language unless the scarcity is mechanically verifiable. It does not respond to social proof unless the proof is statistically valid. The persuasion architecture of consumer marketing for the last twenty years was built for a buyer that the agent simply is not.

The brands losing share to agents are not losing because their product is worse. They are losing because their product is illegible to a machine. The brands winning are the ones that became legible first. Field note, Q1 2026 merchant interviews

The three new evaluation axes.

From our interview set and the structured data we pulled from observable agent purchase patterns, three evaluation axes are doing the heavy lifting in how agents differentiate between competing products. None of them map cleanly to the categories most brand teams measure against today.

Structured trust. The agent prefers brands that publish their specifications, sourcing, returns terms, and performance data in machine-readable form. Schema markup, structured snippets, verifiable certifications. The brand that publishes a clear material composition wins over the brand that publishes a beautiful product story but leaves the material vague. The agent cannot resolve the vague. So it discounts it.

Behavioral consistency. The agent watches what your brand does over time. Price stability matters more than price level within a band. Inventory reliability matters more than inventory depth. Returns experience matters more than returns generosity. The agent is building a probabilistic model of how likely each merchant is to deliver on the implied contract, and it adjusts its recommendation weights accordingly. Brands that run frequent promotional volatility are actively training agents to discount them.

Defensible specificity. The agent rewards brands that are clearly the best at a narrow thing over brands that are broadly competent at many. When the agent is searching for "lightweight cycling jacket for sustained light rain at 12 to 18 degrees," it does not want the omnibus performance brand. It wants the brand whose entire positioning is engineered around exactly that condition. The implication for product line strategy is significant and largely unaddressed.

What this means for brand investment.

If the diagnosis above is correct, the implications for how brand teams allocate budget over the next four to six quarters are uncomfortable. Several large investment categories that have been productive for years are entering a period of diminishing returns. Several smaller categories that have historically been treated as technical hygiene are now becoming primary battlefields.

The categories that are decaying in agent-mediated commerce include performance creative built around aesthetic differentiation, mid-funnel emotional storytelling, influencer placements that are not paired with structured data, and brand-name search ads in categories where the agent has already pre-selected a shortlist. The categories that are appreciating include structured data quality, third-party verification programs, returns infrastructure, on-product specification clarity, and the unglamorous discipline of making your brand legible to the layer of software that is increasingly making the call.

This is a difficult message for most marketing organizations because the appreciating categories sit outside their reporting line. Structured data lives with the e-commerce team. Specification quality lives with product. Returns experience lives with operations. The brand team has historically been the consumer of these inputs, not the owner. That allocation of authority needs to change, or the brand team will continue to be measured against the wrong scoreboard.

The five things to do this quarter.

The harder question.

The harder question is what happens to brand loyalty when the loyalty is held by the agent, not by the human. Today the agent is largely deferential to stated user preference. The user says "buy from Brand X" and the agent buys from Brand X. But the agent is also learning. It is observing which brands actually deliver on the implicit promise, and it is increasingly making proactive recommendations based on its own evolving model of which brands are reliable.

Over the next two to three years, the agent will become the primary holder of brand preference in a meaningful share of categories. The human will set a high-level intent. The agent will execute the choice within that intent. Brand affinity, as we have understood it for the last century, will be partially intermediated by software that has its own evaluation function and its own institutional memory.

This is not a science-fiction scenario. It is a forecast we hold with high confidence over an eighteen-month horizon. The implication for how brands build long-term differentiation is structural. The implication for how marketing organizations are staffed and measured is significant. The implication for the share of budget that should be moving toward the engineering and data side of the brand stack is direct and urgent.

The brands that read this signal early and acted on it in 2024 and 2025 are already pulling ahead in their categories. The brands that read it now and act before the second half of 2026 will still capture significant first-mover advantage. The brands that wait for the trend to be fully visible in the data their existing dashboards report will be late. That is the entire bet.