Alexa+ handles food orders

Amazon has spent more than a year telling people that Alexa+ would be different from the old Alexa. On March 31, that promise turned into a very specific product: Alexa+ can now place food delivery orders from Uber Eats and Grubhub through a back-and-forth conversation instead of the brittle command format that made voice assistants feel like menu trees with a microphone (aboutamazon.com). The change matters because ordering dinner is the kind of task that exposes every weakness in voice software. People hesitate, change cuisines, ask side questions, swap drinks, add dessert, and want the total before they commit. Amazon is saying Alexa+ can now survive that mess. That is the real story here. This is not just another integration. It is Amazon trying to prove that a large language model can do useful work in the physical world without falling apart the moment a user goes off script. When Amazon introduced Alexa+ in February 2025, it said the system had been rebuilt to orchestrate APIs at scale, because language models alone are good at talking but bad at actually completing tasks like bookings, shopping, or orders (aboutamazon.com, aboutamazon.com). Food delivery is a clean test of that claim. The assistant has to keep track of context, map messy speech onto structured restaurant menus, and update a cart without making the user start over. Amazon’s own description of the feature is revealing. The company says users can begin with a vague request like Italian or Mexican food, jump to a specific restaurant, ask questions, change quantities, modify items, and switch direction mid-order while the cart updates on screen in real time (aboutamazon.com). TechCrunch reported that linked Grubhub or Uber Eats accounts also pull in past orders, which makes reordering easier and gives Alexa+ more context to work with (techcrunch.com). That screen is not a side detail. It is the safety rail. Amazon is not trusting voice alone. It is using a display to show the live state of the order so users can catch mistakes before checkout. That hardware constraint gives away how cautious this rollout still is. CNET reported that the feature worked on an Echo Show and that Amazon said customers need an Echo Show 8 or larger, not just the Alexa app or a speaker-only Echo (cnet.com). TechCrunch likewise said the rollout is starting with Alexa+ customers who have Echo Show 8 devices and larger (techcrunch.com). That limitation makes sense. Conversational AI sounds magical right up until it gets your burrito wrong. A visible cart is how Amazon lowers the risk while still claiming the interaction is natural. The business model is just as practical. Alexa+ is now broadly available in the US, included with Prime and otherwise priced at $19.99 a month, which means Amazon finally has a premium assistant that has to justify itself with repeat behavior, not novelty demos (aboutamazon.com, amazon.com). Amazon says customers are already using Alexa+ for more complex tasks and interacting with it more than twice as much as before (aboutamazon.com). Food ordering fits that strategy because it is frequent, measurable, and tied to partner commerce. Amazon can see where people abandon the flow, where they ask for help, and whether a conversation actually ends with pad thai at the door. That is why this feature feels more important than it looks. The old Alexa was built for one-shot requests. Alexa+ is being trained on recovery. If a user says they want burgers, then remembers a vegetarian in the house, then asks what travels well, then decides to reorder last week’s fries from somewhere else, the assistant has to hold the thread together without exposing the machinery underneath. Amazon is betting that the future of voice assistants will not be won by better answers. It will be won by fewer abandoned carts on an Echo Show 8 sitting on a kitchen counter (aboutamazon.com, cnet.com).