You know the feeling. You’re starving, you tell your AI assistant “order me lunch,” and within seconds, a burrito is on its way. But then, a moment later, a cold dread hits: Did I really want a burrito? You check the order. You didn’t even choose the toppings. You instantly regret it—but who do you blame?
Here’s the uncomfortable truth: AI ordering is turning into a blame sponge—and it’s protecting bad restaurants.
We’ve been sold a vision of frictionless convenience. Open your mouth, food arrives. No scrolling, no decision fatigue. But the problem isn’t the technology. It’s the psychology. When you order manually and get a terrible meal, you blacklist the restaurant. You leave a bad review. You tell your friends. The restaurant feels the pain. But when an AI orders for you and the meal is awful, you don’t blame the restaurant—you blame the AI. “Stupid AI, why did it pick this place?”
This isn’t a hypothetical. I’ve seen it firsthand. A friend trained his AI for weeks, gave it full access to his payment and preferences. The first time it tried something new—a “surprise me” command—he got a soggy salad. His immediate reaction? “This AI sucks.” Not “That restaurant sucks.” The restaurant got off scot-free.
Repeat this scenario a hundred times. The AI gets blamed a hundred times. Each restaurant gets blamed only once (if at all). The consequence? Bad restaurants face no negative feedback. They continue serving mediocre food because the AI is the scapegoat. The AI becomes a shield for mediocrity.
Now let’s talk business. Third-party AI companies want to build ordering assistants. But where’s the revenue? If they don’t own the platform, they can’t take a cut of transactions. They’re stuck selling subscriptions or data—neither of which scales easily. And if they do integrate with platforms like Uber Eats, they risk being squeezed. The platform itself might build its own AI—but then it faces the same blame problem. Does the platform want to absorb user frustration when orders go wrong? Probably not.
There’s a deeper tension here. AI is designed to predict your past preferences. But humans are fickle. You might want a change, but the AI doesn’t know that. It optimizes for consistency, not novelty. And when it tries to be novel, it often fails. That’s not a bug—it’s a feature of human unpredictability. The most efficient system is not always the most satisfying one.
So what’s the solution? First, we need to separate the AI’s recommendation from the user’s final decision. The AI should suggest, not take action. Let the user confirm. That adds friction, but it preserves accountability. Second, platforms need to ensure that feedback loops still work—if a restaurant is bad, the user must know it was the restaurant’s fault, not the AI’s. Third, we need to stop pretending that efficiency is the only metric. Trust and accountability matter more.
Here’s my take: AI ordering, as currently envisioned, is dangerous. It undermines the market’s natural quality control mechanism—the angry customer. Without that, bad restaurants thrive, and good ones lose their edge. Don’t let AI become the shield for mediocrity. Demand systems that keep humans in the loop and blame where it belongs.
FAQ
Q: But isn't AI ordering just a tool? Can't users learn to blame the restaurant?
A: In theory, yes. But psychology shows we anthropomorphize AI and treat it as an agent. Once you blame the AI, you stop thinking about the restaurant. The damage is done, and the restaurant gets a free pass.
Q: So what should a product manager do to fix this?
A: Build in a confirmation step. Make the user explicitly choose the restaurant. Never let the AI complete the order without human sign-off. That preserves accountability and keeps the feedback loop intact.
Q: Isn't this a temporary problem that better AI will solve?
A: No. The problem is not AI accuracy—it's human nature. People will always find a scapegoat. Better AI might make fewer mistakes, but the blame dynamic remains. The only fix is changing the system design, not the algorithm.