Blog
Short essays on how AI is changing what gets recommended, and what that means for brands.
10 posts
AI recommendation systems are reintroducing scarcity into digital commerce.
Digital commerce spent years pretending abundance had solved the old shelf-space problem. AI shopping is reversing that logic. The new digital shelf is not an index of everything — it's a constrained answer space where inclusion matters more than rank ever did.
OpenAI, Google, Shopify and Amazon are all fighting for the same choke point.
OpenAI adds shopping. Google builds AI buying tools. Shopify pushes merchants into conversational channels. Amazon experiments with Buy for Me. Different interfaces, same prize: control the route between merchant systems and AI-mediated demand.
In AI commerce, the real advantage is often not brand fame but machine legibility.
LLMs don't encounter brands the way humans do. They work through representations — structured attributes, merchant data, retrievable descriptions, third-party signals. The machine often picks the competitor it can interpret more cleanly, even when yours is the better product.
The companies that matter next may be the ones that make AI visibility measurable.
Every platform shift eventually creates a supporting industry that looks optional, then becomes unavoidable. Search had SEO. Mobile had analytics. AI commerce is reaching that point now — and the vacuum sits around a question most merchants can't yet answer: are we being recommended, or not?
The assistants may get the attention, but standards will decide who can actually transact.
Markets obsess over assistants because assistants are visible. But the deeper power is accumulating one layer below — in the shared standards that determine who can discover, compare, authorize and transact at scale. Agentic commerce is entering that phase now.
When advice becomes computational, discovery becomes infrastructure.
Beauty shoppers rarely start with a SKU. They start with a need-state — a breakout, rosacea, fine hair after pregnancy — and ask for interpretation. That makes beauty native territory for AI assistants, and the first mainstream category where the full logic of AI commerce becomes impossible to ignore.
The real bottleneck is not checkout. It is recommendation quality.
The easy reading of Walmart's AI shopping stumble is that conversational commerce was overpromised. The more useful reading is harsher: AI commerce fails when the recommendation layer doesn't earn enough trust for delegation — and that has little to do with checkout.
Agentic commerce is shifting advantage from web design to system readiness.
For most of ecommerce history, the storefront was the strategic center of gravity. Agentic commerce is moving that weight down the stack — toward feeds, protocols, payments, and whether your systems can be acted on by a machine.
AI recommendation layers are reshaping demand before analytics can record it.
Every performance dashboard tells a story — and most of those stories are starting to lie. AI assistants are shaping the consideration set before a session is recorded, and the merchant's analytics can't see the part of the journey that now decides whether demand arrives at all.
In AI commerce, the real loss is not traffic. It is the right to compete.
Search exposed the market and let the user do the pruning. Selection prunes the market first. The economic shift is not from typing to talking — it is from exposure to eligibility.