Which middlemen — travel agents, brokers, comparison sites — are easiest for AI shopping agents to cut out first?

This explores which kinds of middlemen are most exposed when AI agents start searching, comparing and negotiating for people. The corpus doesn't offer a ranked list of industries. What it does offer is more useful: a way to tell which intermediaries are fragile. The short answer is that intermediaries whose main value is reducing search and communication costs are the most exposed. Intermediaries that hold settlement, liability or data are proving much harder to dislodge.

Start with what agents actually make cheap. Microsoft researchers argue that the deeper economic effect of agents is not automating tasks. It is collapsing the communication friction between consumers and businesses, which lowers switching costs and shifts market power Could agent-to-agent communication reshape market power?. If an intermediary mostly exists because finding, comparing and contacting sellers is tedious, agents attack its reason for existing. Andrew Chen adds a twist: agents don't build network effects of their own, because they can talk to each other through standard APIs Do AI agents actually benefit from network effects?. So the moat that protected many platforms, being where everyone already is, matters less when software can reach everyone directly.

Now the counterexample, which is the most surprising finding here. Travel looks like an obvious target, since online travel agencies are classic search-and-compare middlemen. Yet Goswami's 2026 data shows the reverse. Independent hotels became *more* dependent on OTAs (63.4%), and Expedia's business-to-business segment grew 24%, because AI agents book through the OTAs' APIs. The OTAs remain the merchant of record, meaning the party that takes payment and handles refunds and disputes Are AI agents actually removing online travel agencies from bookings?. The lesson is that the search layer can be replaced while the settlement layer gets stronger. Intermediaries that carry payment, liability or fulfilment become the plumbing agents run on.

Market structure also shapes the answer. A formal model shows that a smart recommendation algorithm beats ordinary price competition only in *thin* markets with few sellers. As sellers multiply, its advantage shrinks toward zero Does algorithmic recommendation lose its edge as sellers compete more?. Choudary warns that agents acting only for buyers change where shoppers enter a market, not how the market works. Real restructuring needs agents on the seller side too, so that both represented sides can negotiate Do one-sided consumer agents actually change market structure?. Put together, an intermediary is weakest when it only routes buyers. It is safer when it organises both sides or controls the terms of the deal.

Two things slow replacement down. The UK competition regulator (the CMA) argues that without data portability, agents can't act effectively for consumers, and lock-in to incumbent ecosystems blocks switching Should data portability be infrastructure for agent-driven markets?. Intermediaries that hoard customer history are therefore protected until rules force that data open. Second, near-free agent search doesn't automatically produce better markets. Look-alike listings and low-quality sellers can leave shoppers worse off Can cheap agent search actually harm shopping markets?. That gives curators and trust brokers a new job rather than ending their old one. And as agents take over choosing, the value may not disappear so much as move: services start competing for the *agent's* attention, which rebuilds ranking and advertising layers aimed at machines Will agents compete for attention just like users do?. Whoever builds those agents decides whose interests they optimise Does agent ownership determine whose interests get optimized?.