THE SHORT VERSION
- Starbucks is preparing to test AI models that run in the background and help baristas input orders more accurately and efficiently, chief executive Brian Niccol said on the company's earnings call.星巴克 CEO 布莱恩·尼科尔在财报电话会上表示，公司准备测试在后台运行、帮助咖啡师更准确高效录入订单的 AI 模型，目标是让员工专注于与顾客对话。
- The chain is also exploring technology that could recognise participating customers as they arrive, present personalised digital menus and make payment easier, which the company says remains early-stage work.公司还在探索识别「参与的顾客」并在其抵达时呈现个性化数字菜单、简化支付的技术；星巴克称这项工作是早期阶段。
- Other chains have tested drive-thru AI for years without mass adoption: Burger King has no plans for a broader rollout of its ordering pilot, while McDonald's says its version could free labour hours.同业已测试得来速 AI 多年但未大规模普及：汉堡王不打算扩大其点单试点，称顾客更愿意与人交流；麦当劳称其版本每周可为每家门店节省数十个工时。
Background models, not a robot voice后台模型，不是机器嗓
Starbucks is preparing to test artificial intelligence models that operate in the background to help baristas input orders more accurately and efficiently as they interact with customers chief executive Brian Niccol said on an analyst call, according to Bloomberg's report published by The Hindu BusinessLine. The stated goal is for workers to keep talking to customers while the system handles part of the order entry.据彭博社报道（由《印度教徒商业在线》刊载），星巴克 CEO 布莱恩·尼科尔在分析师电话会上表示，公司准备测试在后台运行的人工智能模型，帮助咖啡师「在与顾客互动时更准确、更高效地录入订单」。公开目标是让员工持续与顾客对话，由系统承担部分订单录入。
That is a narrower proposition than the drive-thru voice assistants that have been piloted across the industry. The model is not being positioned as a replacement for the conversation; it sits behind the counter, on the employee's side of the transaction, and its first job is accuracy rather than autonomy.这比同业试过的得来速语音助手更窄。该模型不被定位为对话的替代品，而是位于柜台之后、员工一侧，首要任务是准确，而不是自主。
Menus that recognise the car认得出车牌的菜单
The second part of the plan is more visible to customers. Niccol said the company would explore experiences that recognise participating customers as they arrive, present personalized digital menus and make payment more effortless according to The Verge. The word participating matters: the recognition is opt-in, not automatic identification of anyone who pulls up.计划的第二部分对顾客更可见。据 The Verge 报道，尼科尔表示公司将探索「在参与的顾客抵达时识别他们、呈现个性化数字菜单并让支付更省力」的体验。「参与」这一表述很重要：识别是选择加入的，而不是自动认出所有驶入车道的顾客。
The mechanics are not public. The Verge notes that the company has not said how soon personalised digital menu boards might appear, and Niccol described the work as “early”, adding that the company will “test, learn, and scale what works.” A menu that changes per car also raises questions the call did not answer, including what a customer sees when recognition fails.具体机制尚未公开。The Verge 指出，公司未说明个性化数字菜单板何时可能出现，尼科尔称这项工作是「早期」，并会「测试、学习，然后推广行之有效的部分」。随车变化的菜单也带来电话会上没有回答的问题，包括识别失败时顾客会看到什么。
2027 is the date on the record记录在案的时间点是 2027 年
One concrete timing detail did surface. Starting in 2027, Starbucks will begin testing background AI models at the drive-thru, according to The Verge's account of the call. The same report says the company plans to integrate AI into the app's search and discovery features, and may allow “conversational and agentic” ordering later.有一项具体时间信息浮出水面。据 The Verge 对电话会的报道，自 2027 年起，星巴克将开始在得来速测试后台 AI 模型。同一报道称，公司计划把 AI 集成到应用内的搜索与发现功能，未来可能允许「对话式与智能体式」点单。
Those are separate programmes with separate risk profiles. Helping a barista enter an order is an internal workflow change; letting an agent order on a customer's behalf touches payment, consent and error handling. The company has described the first in more detail than the second.这些是风险特征不同的独立项目。帮助咖啡师录入订单属于内部流程变化；让智能体代表顾客下单则涉及支付、同意与错误处理。公司对前者的描述比对后者更具体。
The pieces already in production已经上线的部分
Starbucks has already shipped some AI features. Customers can place orders through ChatGPT, and employees have an AI-powered tool called Green Dot Assist that they can ask about recipes and service standards, according to The Verge. Those are live products rather than tests, which makes the drive-thru work an extension of an existing programme.星巴克已上线部分 AI 功能。据 The Verge 报道，顾客已可通过 ChatGPT 下单，员工则可使用名为 Green Dot Assist 的 AI 工具询问配方与服务标准。这些是正式产品而非测试，因此得来速项目是既有计划的延伸。
The pattern is consistent across the industry: assistants that answer questions ship first, because a wrong answer is cheap to correct. Agents that take actions ship later, because a wrong action costs money. The drive-thru plan sits on the boundary between those two categories.整个行业的路径也一致：回答问题型助手先上线，因为答错的成本低；执行动作型智能体后上线，因为做错的成本高。得来速计划正好位于这两类之间。
Why the industry has not scaled this yet为什么行业还没规模化
Large chains have tested AI in the drive-thru for years, Bloomberg notes, but it has not reached mass adoption. Burger King has no plans for a broader rollout of the AI ordering system it is piloting, saying customers prefer to talk to humans. McDonald's is testing its own version and says it could free up dozens of labour hours per restaurant each week.彭博社指出，大型连锁已测试得来速 AI 多年，但尚未大规模普及。汉堡王不打算扩大其正在试点的 AI 点单系统，称顾客更喜欢与人交流；麦当劳正在测试自己的版本，并称每周可为每家门店节省数十个工时。
That split is the honest summary of the evidence so far. The technology can work; whether customers want it is a separate question from whether it saves time, and the two answers can point in opposite directions for different brands.这种分化才是目前证据的诚实总结。技术可行；但顾客是否想要，与是否省时是两个不同问题，而对不同品牌，两者的答案可能相反。
Why this quarter mattered for the pitch为什么这季度对这套说法重要
The AI plans were discussed alongside earnings. Starbucks reported results that surpassed expectations on Thursday, as a plan focused on restaurant upgrades, hospitality and new menu items continued to drive results, according to Bloomberg.AI 计划是在财报背景下讨论的。据彭博社报道，星巴克周四公布的业绩超出预期，原因是以门店升级、服务体验与新菜单为重点的计划持续带来成效。
That context changes how the AI announcement should be read. A company whose turnaround is working can present automation as an addition to service rather than a substitute for it; a company under pressure would have a harder time making the same case to baristas and customers.这一背景改变了 AI 公告的读法。转型奏效的公司可以把自动化描述为服务的补充而非替代；若公司承压，向咖啡师与顾客讲同样的故事会难得多。
What would show the plan is real什么能证明计划是真的
Three things would move this from announcement to evidence: a named test market and a date, disclosure of how customers opt in to recognition, and a measured result for order accuracy or speed. None of those is in the call material reviewed here.三件事能把「公告」变成「证据」：具名的测试市场与日期、顾客如何选择加入识别的说明、以及订单准确率或速度的实测结果。已读电话会材料中都没有这些。
Related coverage on this site includes OpenAI's always-on agent and its permission model, Muse for Small Business connecting an assistant to a shop's accounts and Google's small-business Gemini workbench. All three describe the same shift: assistants are moving from answering to acting, and the controls are the part that decides whether that is useful.本站相关报道包括 OpenAI 常驻智能体与其权限模型、Muse 小企业版把助手接入店铺账户与谷歌面向小企业的 Gemini 工作台。三者描述同一种转变：助手正从「回答」走向「执行」，而控制机制决定这是否有用。
What the rivals have already tried同行已经试过什么
Starbucks is not the first chain to put a changing menu in front of a car. The Verge reports that Taco Bell is experimenting with technology that could change the layout, content and visuals of a digital menu board “on a car-by-car basis”, and that McDonald’s has shown off an AI feature called ArchIQ that identifies repeat customers and lets them order their usual.星巴克并非第一家把「会变的菜单」放到车前的连锁。据 The Verge 报道，塔可钟正在试验可在「逐车」基础上改变数字菜单板布局、内容与视觉效果的技术；麦当劳则展示了名为 ArchIQ 的 AI 功能，可识别回头客并让他们点「老样子」。
Those examples describe three different bets. A menu that reorders items is a design problem; a system that recognises a returning customer is a data and consent problem; a model that helps an employee type faster is an operations problem. Starbucks is working on all three at once, which is why its own description stays at the level of what it will explore.这些例子代表三种不同的押注：菜单重排是设计问题；识别回头客是数据与同意问题；帮员工更快录入是运营问题。星巴克同时在推进三者，这也解释了为何其官方描述停留在「将探索什么」的层面。
Sources & reporting limits来源与核查范围
Follow Industry to find related stories on your next visit.关注产业动态，下次访问时阅读相关文章。
