Starbucks just quietly killed an AI inventory system it deployed across 11,000 stores, nine months after the launch announcement, and five months after employees started complaining it couldn’t tell oat milk from 2%. That’s not just a Starbucks problem. Enterprise AI is failing at staggering rates right now, and the reasons are almost always the same. Today the win is a lesson.
We break down exactly what went wrong with NomadGo’s Inventory AI, why workers ended up doing twice the work instead of half, what the deleted press release tells us about how these rollouts get sold internally, and the pattern that shows up in almost every failed enterprise AI deployment, from coffee shops to pizza chains.
You must log in or # to comment.
no one told them llms cant count? computer vision can. hilarious how no one knows that llms arent the only ai
Welcome to the wonderful world of “context”. In conversation, context is the absolute ruler. Failure to follow context just makes the person doing so look foolish.
In modern contexts, AI, a diverse field that has about three quarters of a century behind it now, with many branches of its technology, refers, unless you’re in only a technical context with fellow practitioners of art, almost exclusively to the (de)generative AI that is being crammed into our collective orifices from all angles. Older forms of AI, the ones that have each had their bubbles of varying sizes, subsequent collapses, until they found their little niche where they fit perfectly, aren’t generally called AI any longer. They’re called “software”.
Like my phone, which predates the current round of LLM slop mongers, has camera filters and capabilities that were at the time touted as “AI” and that features prominently in my phone’s camera’s user interface. Later phones from the same company with the same capabilities (and then some) don’t have the “AI” branding because nobody thinks of this as AI any longer. They think of it as “normal camera software”.
So in this real world context, unless you’re speaking strictly in a professional capacity to practitioners of the art, the term “AI” refers, in common conversation, to LLMs and other forms of (de)generative AI 99.44% of the time. Anybody who professes not to understand this is either an asshole or a bonehead.
Why does this summary read just like AI?
Reads like generic video essay summary to me. If AI output sounds like that, it’s probably because AI was trained on that type of writing.
