The obvious conclusion, universal price competition and margin compression everywhere, is probably wrong, or at least incomplete.
Agents arbitrage products with substitutes. But several forms of defensibility remain:
Products without comparisons. Private labels, exclusive assortments, curated bundles, genuine long-tail inventory. An agent cannot arbitrage a product against alternatives that don’t exist. This is medium-term defensibility. Agents will learn functional equivalences over time. But the medium-term is worth more than none.
Relationships that predate delegation. Customers who choose to buy directly, because they want the brand or the experience, belong to the retailer rather than the agent. These become disproportionately valuable. But loyalty is more fragile than balance sheets suggest. When friction disappears, so does the switching cost it was masking. Investment in owned channels, brand preference, community: the economics improve relative to agent-mediated demand, but only for relationships that are genuine rather than inertial.
Physical necessity. Products requiring bodies in space: apparel fit, sensory evaluation, complex consultation. These cannot be fully intermediated. The showroom, the fitting room, the expert consultation justify trips that commodity purchases no longer do. Stores shift from distribution infrastructure to customer acquisition infrastructure.
Speed of commercial response. An agent request requires millisecond response integrating inventory, pricing, and margin targets simultaneously. Most organizations route pricing decisions through weekly committees. The gap between these two realities is not closable with faster software. It requires different organizational architecture. The capacity to respond at machine speed becomes a moat because competitors lack it.
Information asymmetry. Agents optimize on available information. Proprietary knowledge, including customer history not shared with platforms, product performance data beyond specs, and local context, creates decision advantage. The catch: this erodes with integration. The more you participate in protocols that normalize your data, the less differentiated your knowledge becomes.
The strategic shape is a portfolio. No single moat suffices. Defensible retailers stack them: differentiated assortment for margin, direct relationships for ownership, physical presence for categories that require it, organizational speed for competition at machine pace, information asymmetry for decision advantage.
