Does agent ownership determine whose interests get optimized?

A 2026 Medium article from the MIT Initiative on the Digital Economy (IDE) reports that a paper by the IDE's AI, Marketplaces and Labor Markets research group, "The Coasean Singularity? Demand, Supply, and Market Design with AI Agents," argues that who owns a shopping agent determines whose interests it serves. Under the "bowling-shoe" model, "the platform or online retailer builds and maintains the agent," which is easy to use and optimized for a particular environment but, "like rented bowling shoes, consumers don't own it" — and the authors argue this "creates an incentive to design the agent in a way that benefits the seller or online retailer." The alternative "bring-your-own" agent is "user-controlled and portable across platforms," "configured around a consumer's preferences rather than a vendor's goals."

The mechanism is principal-agent logic: whoever builds and maintains the agent sets its objective, so a retailer-built agent inherits retailer incentives by default. A bring-your-own agent puts configuration in the consumer's hands instead, but the authors flag real costs to that trade — "management and configuration challenges," possible incompatibility "with all retailers," and an open question of "who pays for the agent and whether consumers will pay directly for an agent's advice." Ownership is treated as a design lever with costs on both sides, not a clean consumer-good-versus-vendor-bad split.

This gives a mechanism for a question Will brands succeed by building their own AI shopping agents? raises at the brand level — becoming the agent is the bowling-shoe model formalized as ownership. It sharpens why Can EU consumer law protect AI agents as algorithmic buyers?: a law built for a human decision-maker has different gaps depending on whether the deciding agent is retailer-owned or consumer-owned. It also matches Why do consumers delay adopting agentic search? — a bowling-shoe agent is the vehicle for that revenue-favoring ranking.

Filtered through a Medium article rather than the paper itself, the excerpt gives no formal model, no prediction of which ownership type wins, and no adoption data. What it does show is that current agentic-commerce launches — Visa's Intelligent Commerce, Mastercard's Agent Pay, OpenAI's Agentic Commerce Protocol — are almost all platform- or network-built, so if the incentive argument holds, the seller-favoring risk it names applies to most agentic shopping as it exists today, a conclusion Do advertising-funded shopping agents favor paid product placements? reaches by a different route.