This letter is addressed to the teams who built the platforms in the restaurant AI category — the decision intelligence products, the observability tools, the management suites, the benchmarking platforms, the AI assistants.
You built something real. You talked to operators. You understood the pain. You raised capital and hired engineers and shipped product and acquired customers. Most of you are still growing. Most of your customers are still paying.
You deserve a clear explanation of what changed.
What You Built
You built the right answer to the wrong question. The question you answered — how do we give restaurant operators better information? — is a legitimate question with a real market. The dashboards you built are accurate. The alerts you surface are relevant. The benchmarks you provide are genuinely useful.
The question that matters is different: what happens between the moment a signal becomes detectable and the moment an operator can act on it?
Your answer to that question, built into every architecture in this market, is: a human reads the alert and acts. The human is the bridge between intelligence and intervention.
That human is on the floor. She is managing a team. She has seven things happening and is being interrupted by an eighth. The alert arrives. She sees it. She is already somewhere else. The window closes. The guest decides.
That is not a UX problem. It is not a notification design problem. It is an architectural premise: you built for the operator who has time to read, and deployed into operations where no such operator exists.
Why This Is Not a Roadmap Problem
We anticipate that some of you are already scoping the response. A new feature. A faster alert. An AI layer that takes some action before the human has to. A partnership that adds execution capability to your awareness infrastructure.
These are understandable responses. They are not sufficient ones.
The behavioral corpus that enables genuine execution-layer intelligence is not a dataset you can license or synthesize. It was built across fifteen years of venues far larger than a restaurant — stadiums, theme parks, mass retail environments processing 80,000 concurrent human beings. The crowd contagion science embedded in our decisioning engine does not exist in restaurant data because it was never observable at restaurant scale.
You can hire engineers. You can build a streaming pipeline. You can replace Snowflake with Kafka. You can close the latency gap from fifteen minutes to one second. None of that closes the training data gap. None of that gives you the trillions of data elements and six hundred million real purchasing decisions that produced the models that understand what a dining room is actually doing before the POS or the WiFi logs have time to process.
Architecture is not a sprint. Training data is not a sprint. The gap is structural.
What Is Already Happening
The operators who have deployed superGM.ai are not canceling your contracts immediately. Many of them still use your products for Monday morning reporting. What they are not doing is treating your alerts as operational intelligence. What they are not doing is opening your dashboards during service. What they are not doing is routing decisions through your platforms in the moment those decisions need to be made.
They have something that acts on those moments. They will not tell you what it is. They asked us not to tell you either.
You will see it in their renewal conversations. You will see it when their numbers move in directions you cannot explain with data you have access to. You will build a roadmap to close the gap and find that the gap has moved.
What We Are Not Saying
We are not saying you failed your customers. You gave them the best that the category could offer, and the category offered a great deal. We are saying the category has a ceiling, and the ceiling is not a feature — it is the premise that a human is the last mile of every intervention.
We are not gloating. The operators your platforms serve are the same operators we are trying to serve. The hospitality they are delivering — or failing to deliver — affects real guests having real experiences in real rooms. That matters regardless of what stack is running underneath.
We are saying: the premise has changed. The layer you built for was the right layer for 2020. The layer that matters in 2026 is the one your architecture does not reach.
Build toward it if you can. We do not expect you to close it on a roadmap. We wanted to tell you clearly, before your renewal conversations begin, what is on the other side of the numbers that are going to start moving.
— superGM Intelligence Team