An AI agent tried to guess what wine I was drinking based on my description — and the results were mixed to say the least
Salesforce's tAIster AI agent tried hard, but isn't a certified wine snob just yet.
In the realm of AI's expanding influence, a recent experiment at Salesforce's Dreamforce 2026 showcased an AI agent's attempt to identify wines based on the user's descriptions. The session, conducted in a secluded executive lounge, involved participants tasting three mystery wines via an AI agent named tAIster. This AI agent had been trained on an extensive database of over 400,000 distinct wines, encompassing various attributes such as taste, aroma, and texture.
To begin, tAIster prompted users to provide descriptive keywords regarding the appearance and grip of the wine in the glass. Following this, participants were asked to articulate the wine's aroma and finally, its taste. The goal was to provide as comprehensive a description as possible to enable the AI agent to accurately identify the wine.
During this exercise, I, as a participant, attempted to be descriptive in my assessments. My first two wine samples were correctly identified by the AI agent, but the third one proved challenging - it incorrectly presumed the wine to be Veuve Cliquot champagne, rather than the actual Chardonnay, due to my mention of small bubbles in the glass.
Despite this anomaly, the experiment provided an intriguing glimpse into the potential of AI agents in unconventional domains like wine tasting. However, it's evident that real-world sommeliers will still remain indispensable for the foreseeable future.
Written by urgent.news from TechRadar's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
