Why would a store want its own AI shopping assistant instead of letting any outside AI agent in to shop?

This explores why retailers might want AI shopping to run through their own closed systems rather than open ones that any agent can enter. The corpus doesn't have a study that asks retailers about this directly. It does have enough on who gains and loses in agentic commerce to piece together the motive, and the motive is mostly defensive. Waller expects agentic commerce to split the way payment networks did: open systems that accept many agents, and closed ones that require a proprietary agent. That choice will shape how concentrated the market becomes and who owns each layer of it Will agentic commerce platforms adopt open or closed models?. For retailers, the real stake is whether they keep the customer relationship or hand it to someone else's agent.

The fear has numbers behind it. BCG's survey of 2,500 US consumers projects that $1.3 trillion of e-commerce spending will move to agent-assisted or fully autonomous shopping. It finds that consumers and platform networks capture most of the gains while merchants are left at a structural disadvantage Who wins and loses as AI agents reshape shopping?. In an open system, an outside agent compares a retailer's products against everyone else's, and the retailer becomes interchangeable inventory. Choudary explains why this is lopsided: an agent that represents only the buyer changes where shoppers enter the market but leaves the retailer with no agent of its own to negotiate back Do one-sided consumer agents actually change market structure?. A closed system is one way for the seller to keep a seat at the table. Accenture adds that consumers say they prefer agents built by retailers, though that comes from surveys, not from agents shown working at scale Will brands succeed by building their own AI shopping agents?.

The less obvious reason is economic. A formal model finds that an optimal recommendation algorithm beats ordinary price competition mainly in thin markets, where it can hold back some trades. That edge shrinks steadily as more sellers compete Does algorithmic recommendation lose its edge as sellers compete more?. A closed system keeps the market thin on purpose. Whoever runs it controls the ranking, and two other findings show why that control is worth having. One model shows platforms judge agentic search by conversion revenue alone and want it adopted before consumers would rationally switch Why do consumers delay adopting agentic search?. A simulation shows that when a platform tunes its ranking and sellers tune their product descriptions to exploit the same agent biases, harm to consumers more than doubles, with no coordination needed Can platform ranking and seller descriptions separately harm AI shoppers?. So the control that makes closed systems attractive to retailers is the same thing that can work against shoppers.

There is also a practical reason: fraud. Boboev describes the emerging agent protocols as a ladder of layers covering identity, discovery, governance, and the application itself. Each layer absorbs a different level of fraud risk Do agentic commerce protocols compete or complement each other?. When anonymous outside agents can transact, the merchant has to verify all of them. A closed system collapses that problem into a single trusted agent.

The counterweight is regulatory and structural. The UK's competition regulator, the CMA, treats data portability as basic market infrastructure. Without it, agents can't act for consumers and switching stalls under ecosystem lock-in Should data portability be infrastructure for agent-driven markets?. Rothschild and colleagues warn of a further twist: the walled gardens most likely to win may belong not to retailers but to Apple, Google, Meta, and Microsoft, who could rebuild app-store gatekeeping by controlling the layers that let agents work together Will agent economies develop as open webs or closed gardens?. A retailer that goes closed to avoid being commoditized by open agents may end up dependent on a bigger platform's closed system instead.