LLMs

Ordering with the Starbucks ChatGPT app was a true coffee nightmare

Published byAIDaily Editorial Team
5 min read
Original source author: David Pierce

Venti iced coffee, light skim milk. That's what I get at Starbucks. It is what I have gotten at Starbucks every time I've been to Starbucks for as long as I can remember, other than a brief love affair with the caffe misto a few years ago. In person, my brain barely needs to activate […]

Share:

Ordering coffee is easy. Besting the Starbucks app with AI chat is going to be very, very hard.

Ordering coffee is easy. Besting the Starbucks app with AI chat is going to be very, very hard.

Venti iced coffee, light skim milk. That’s what I get at Starbucks. It is what I have gotten at Starbucks every time I’ve been to Starbucks for as long as I can remember, other than a brief love affair with the caffe misto a few years ago. In person, my brain barely needs to activate to say the words aloud; in the app, it’s four taps and I’m ready to go.

My first time ordering Starbucks through its new ChatGPT integration, which launched last week , was comparatively a complete mess. Getting started is easy enough, if not exactly obvious: Just open ChatGPT and type “@Starbucks” plus your order. You can probably guess what happens next, right? I promise you’re wrong. “Order me a Venti iced coffee with light skim milk,” I typed, to which ChatGPT responded: “The Iced Coffee is exactly what you’re after—cold-brewed and served unsweetened, so adding light skim milk will keep it smooth without getting heavy.” Cool, thanks for the info ChatGPT. Please order me coffee.

Above the message, ChatGPT added what I figured out was a menu, showing the three most likely things I might have meant by “iced coffee.” Iced Coffee was the first option, victory! But I had to select “Customize,” then scroll through the pop-up UI and select both the right size and the milk addition, or else when I tapped “Add to cart” I got just a Grande black iced coffee.

I should note that this had already taken longer than it takes to open the Starbucks app, tap “Order,” tap the name of the closest store, tap the plus sign next to the drink I always get, and check out. But I soldiered on: I got the drink I wanted in the cart, and then went to add my wife’s drink to my order. She calls it “the fruity tea,” which is not a name, but is the kind of fuzzy search ChatGPT ought to handle well! It offered me Iced Green Tea Lemonade, which is a reasonable but wrong guess. I eventually remembered it was the Passion Tango Tea, at which point ChatGPT offered me another enthusiastic description of the tea. Once again, I scrolled up, I customized, and I added to cart.

At that moment, I got an ominous pop-up: “This chat is nearing its limit.” I’m a free-tier ChatGPT user, but I haven’t touched the app in weeks (I’m mostly a Claude guy these days), so hitting the limit this fast was a bit surprising. Also, why is there a limit at all, when I’m trying to do a thing that theoretically makes both ChatGPT and Starbucks a bunch of money? To get things done as quickly as possible, I went to check out. Turns out, ChatGPT has my location wrong, and offered a list of stores half a state away from me. When I went to the map view, where ChatGPT said I could change my location, all I got was an “Oops! Something went wrong.” message. And right about then, I got another pop-up: “You’re out of messages with the most advanced Free model.” It told me it would reset — in five hours. Until then, I’d be shunted to some other, lesser model.

Any rational person would have given up a while ago, right? This is a straight-up terrible ordering experience, made vastly more complicated by the back-and-forth chat system that conferred exactly zero discernable AI upside. But, like a good journalist, I tried again — I started over, @-mentioned Starbucks, and told it my order as succinctly as possible. It confirmed my request, and then let me down gently. “I can’t place your order directly or add it to a real cart,” it said, before offering to walk me through how to use the Starbucks app. Evidently, the model I’d been downgraded to didn’t support the more advanced Starbucks features — or have any idea what I’d just been up to.

I can’t shake the idea that this app — like so many AI tools — appears to be designed for people that simply don’t exist. In Starbucks’ own blog post , it suggests you might prompt the app with things like “Recommend a drink that matches the vibe of my outfit” or “I’m in the mood for something cozy and nutty.” Is that how anyone actually decides their beverage of choice? At best, these features are silly fun. At worst, they’ll lead to even more people dreaming up ridiculous, 12-ingredient, made-to-be-TikToked drinks that drive baristas batty all day.

The actual dream of AI coffee ordering has been the same for a long time: I want to say “order me coffee,” and my assistant should know exactly what to get me and from where. The tech industry tried this in the era of Google Assistant and Alexa, and they’re trying again in the times of ChatGPT. There’s a chance that truly useful AI agents, like the ones Google is testing with Gemini , can go click around for you and get the job done automatically. But chat ain’t it, friends. Coffee ordering, like so many things in life, is not a creative experience designed for conversation. It is a transaction. Ideally, a very short one, because I haven’t had my coffee yet.

Silicon Valley has forgotten what normal people want

Huawei beats Samsung and Apple to market with the first wide foldable

This pasta sauce wants to record your family

Read Tim Cook’s letter to the Apple world as he departs as CEO

Key takeaways

  • The integration of ChatGPT with the Starbucks app illustrates the challenges of AI adoption in services.
  • User experience is crucial for the acceptance of digital technologies, especially in markets with a strong culture of in-person service.
  • User-centered design must be a priority to avoid frustrations that could harm the perception of technology.

Editorial analysis

The integration of ChatGPT with the Starbucks app highlights a significant challenge in the adoption of AI technologies in the service sector. While the promise of more personalized and efficient service is appealing, the user experience can be compromised by complex interfaces and interactions that do not meet expectations. For the Brazilian market, where coffee culture is deeply rooted, ease of use is crucial. Starbucks, which already has a loyal customer base, may face resistance if the digital experience is not as seamless as in-person service.

Moreover, this situation underscores the importance of user-centered design in AI applications. User frustration when trying to place a simple order can lead to a negative perception of the technology, impacting the acceptance of similar solutions in other sectors. In Brazil, where digitalization is on the rise, it is vital for companies to learn from these mistakes to prevent technology from becoming a barrier rather than a solution.

What to watch for next is how Starbucks and other tech companies will respond to this feedback. Improvements in interface design and AI comprehension capabilities are essential to ensure that users not only adopt the technology but also view it as a useful tool. Additionally, it will be interesting to see how competitors react to this integration, especially in a market where customer experience is a competitive differentiator.

Finally, the user experience with ChatGPT at Starbucks could serve as a case study for other companies looking to integrate AI into their services. The lesson here is clear: technology should simplify and enhance the customer experience, not complicate it. The success of AI in the service sector will depend on companies' ability to listen to their customers and adapt their solutions to effectively meet their needs.

What this coverage includes

  • Clear source attribution and link to the original publication.
  • Editorial framing about relevance, impact, and likely next developments.
  • Review for readability, context, and duplication before publication.

Original source:

The Verge AI

About this article

This article was curated and published by AIDaily as part of our editorial coverage of artificial intelligence developments. The content is based on the original source cited below, enriched with editorial context and analysis. Automated tools may assist with translation and initial structuring, but publication decisions, factual review, and contextual framing remain editorial responsibilities.

Learn more about our editorial process