Context Engineering Is the Brand-Truth Layer by Another Name

Every year, Scott Brinker tallies the entire marketing technology landscape. He has charted it since 2011, when the field held roughly 150 products. This year's census put the number past 15,000, and for the first time in fifteen years the growth essentially flattened. Peak martech, perhaps. So when the analyst Ad Age once dubbed the godfather of martech picks a word of the year, the selection carries genuine weight. For 2026 he chose "context," and he gave the practice it creates a name: context engineering.

Read his definition with a product catalog in mind. The work, Brinker writes, is "the technical and operational work of making the right data, content, permissions, instructions, models, workflows, and guardrails available to the right agent at the right time." Set the martech vocabulary aside, and that sentence describes, almost line for line, the layer a retail brand needs between its catalog and a shopping agent. The field just handed the thing a name.

From apps you operate to infrastructure agents use

Brinker's broader argument is that the software itself is fundamentally changing shape. Martech platforms, he writes, are "starting to metamorphose from apps humans operate into infrastructure agents can use." For two decades the work meant logging into dashboards and navigating through configurable workflows, with the interface as the product. That premise is dissolving, and his line for where it lands is blunt: "In the last era, martech platforms competed to be where marketers worked. In the next era, they'll compete to be what agents can work with."

Retail has its own version of that proposition, and Brinker points directly at it. He observes that the website's job is being renegotiated as AI assistants and shopping agents become "front-line intermediaries," and in the same breath he calls them "first-class audiences." Web experience platforms and ecommerce carts were two of his fastest-growing categories for exactly this reason. The storefront a shopper browses is one audience. The agent is another, and it never logs in or admires the homepage. It queries the catalog as infrastructure, the way one application calls another. A Shopify store can be wide open for business to humans and effectively invisible to that reader.

What context engineering looks like for a catalog

Brinker splits the idea into three kinds: customer, company, and systems. The first covers the shopper's situation and intent. The systems strand is whatever your stack can actually reach and deliver. The company strand matters most to a seller, and he frames it as "your goals, strategy, brand, processes, capabilities, governance," the things you know about how you operate and what you stand behind.

For a retailer, that company knowledge is the merchant's own hard-won judgment. Which item to put forward when the first choice sells out, who a given piece actually flatters, what the return policy genuinely allows, and how the brand should sound while it is selling. That judgment is exactly the authored reasoning a product feed has no field to hold. Turning it into something an agent can read and act on is the whole assignment, and it is precisely the work Brinker is naming. He even articulates the payoff structure: "Value engineering identifies the value. Context engineering makes it actionable." A worked example of that authored layer on a single product page demonstrates what "actionable" looks like in practice.

What a bare feed leaves out

Brinker's name for the goal is "Golden Context," the point where customer, company, and systems all converge, "immediate enough to be relevant, grounded enough to be trusted." A standard product feed delivers one thin slice of that, the systems one: price, availability, variants, images. It carries none of the merchant's judgment. An agent that knows your stock level but nothing of your reasoning, Brinker warns, lands at "just fast," quick to make a call and far too ungrounded to make it the right one.

Substitution is where that gap opens first and cuts deepest. When a shopper's first pick is gone, something has to choose the replacement, and a feed that reports only "out of stock" hands that choice straight to the model. How these systems actually reach a decision governs what fills the silence, and what they need from a catalog to choose well is the company judgment a bare feed leaves on the floor.

Naming a layer is how a field agrees it is real

Let me be precise about the limits of Brinker's claim. His report covers marketing operations rather than retail merchandising, and the translation to catalogs and shopping agents is mine. He does reach toward commerce on his own, though, naming shopping agents directly and listing "what inventory we actually have" among the context that matters, so the bridge is short and well lit.

What changes when an industry names something is that the thing stops being one company's pet theory and becomes a shared discipline with a job title attached. Brinker has marketing ops moving "from system admin to stack wrangler to context engineer." Retail has a more concrete name for the same destination: the brand-truth layer between a catalog and the agents reading it. Engineer that context deliberately, or surrender it to whatever the model improvises on your behalf. The agents are already querying. They will act on whatever the catalog gives them, and the brand that engineered its context is the one they can work with.

Sumit Jagdale is the founder of Sartorial.