The closed stacks only win if the rest of the market lets them.

In June 2026 American Express agreed to buy TheFork, and DoorDash launched an AI assistant on top of SevenRooms and Deliveroo.

Within days of each other, American Express agreed to buy TheFork and DoorDash launched an AI assistant that picks where you eat and order from, sitting on top of SevenRooms and Deliveroo, both of which it owns. Two very different companies, the same move: get between the guest and the restaurant and own the data that flows through.

The question is what everyone else does now. The closed stacks only win if the rest of the market lets them. If you run an independent platform or a POS, or a venue wanting to keep its options open, the moment you need to connect to one of these stacks, the stack decides whether you can. It owns the integration and the data, and its incentive is to keep both shut, because the closed loop is the product. You are not a partner in that. You are a dependency.

Hotels have shown a way out. In 2002 a group of hotel companies and vendors set up Hotel Technology Next Generation to build open standards so their systems could actually talk to each other. It worked and it lasted. But when HTNG shipped its Express PMS API in 2022, twenty years on, it still called connecting hotel systems one of the industry’s toughest problems.

The lesson is not that standards fail. It is that a standard fixes the technical layer and leaves the commercial one alone. The connection becomes possible, and the dominant vendor still decides who uses it, on what terms and at what price. Being open was never in their interest, so they adopt it where it suits them and gate it where it doesn’t. Restaurants never even built the equivalent, which is part of why restaurant consolidation is running with nothing in its way.

So the answer isn’t one neutral connector in the middle. A single intermediary, however independent it claims to be, is still one company that can be bought or routed around. The version that works is an open standard the independent middle builds together, reservation platforms like Zenchef, TableCheck and Eat App, POS players like Toast and Lightspeed Commerce. Not a way into the walled gardens, those stay shut, but a parallel pole, good enough to give a locked-in venue somewhere to defect to and keep everyone else from getting trapped.

None of it gets built unless operators ask for it, and we have seen this before. Everyone watched the OTAs take the guest relationship for fifteen years and almost nobody organised, because each venue on its own still needed the bookings that night.

It won’t be every venue that moves first. The leverage sits with the brands and groups whose guests come looking for them by name, the ones who can credibly walk. The rest follow once it’s proven.

This time the alternative is cheaper to build than ever and the AI booking channel is still forming, so there is a window. But a window is all it is. The lock-in isn’t inevitable. It just hardens by default unless operators decide they want the choice and back the people building it. Last time the bill arrived years later. This time you can see it coming.