The sommelier is one of the few roles in hospitality that has historically been considered genuinely irreplaceable by technology. The knowledge base — spanning thousands of producers, regions, vintages, varietals and their interactions with specific dishes — the service skills, the reading of a table's preferences from minimal cues, the ability to make a recommendation that feels personal rather than algorithmic: these have seemed, until recently, resistant to the kind of systematisation that has automated large parts of the reservation, ordering and payments process.
The past two years have produced the first wave of AI-powered wine tools that are serious enough to change that assessment, if not to reverse it entirely. Several of them are already deployed in UK restaurants and hotels; more are in development or in pilot. The honest evaluation of what they can and cannot do is more interesting than either the enthusiasm of their developers or the scepticism of the sommeliers they are described as replacing.
What the Tools Do
The leading AI wine recommendation platforms — including Pix from Wine Access, Preferabli, the wine modules embedded in OpenTable's newer features, and several UK-developed tools including Vivino's restaurant integration — operate on broadly similar principles. A large training database of wines, their flavour profiles, regional and varietal characteristics, critic scores and user review data provides the knowledge base. A recommendation engine, trained on pairing data and user preference feedback, suggests wines from the operator's list in response to a dish selection, a stated preference or a price constraint.
The output, presented via a QR code at the table, a tablet interface or integration with the menu management system, gives the diner a wine recommendation without waiting for a sommelier's attention.
The more sophisticated platforms go beyond simple pairing matching. Preferabli's system, for example, learns individual user preferences over time across multiple interactions, building a taste profile that makes subsequent recommendations more personally calibrated. The OpenTable wine feature uses table booking data and previous visit records to surface recommendations that reflect the guest's ordering history.
Where AI Genuinely Adds Value
The honest assessment from operators who have deployed wine recommendation AI alongside human sommeliers — rather than instead of them — is that the technology adds genuine value in specific contexts that have nothing to do with replacing expert knowledge.
The first is scale and availability. A restaurant with 200 covers and one sommelier cannot have a wine conversation with every table at the moment it is needed. An AI tool available to every diner simultaneously, through their own device or a table-mounted interface, extends wine guidance to tables that would otherwise receive none — which, in most restaurants, is the majority of tables on a busy service.
The second is the information gap for less wine-engaged diners. A guest who genuinely does not know where to start with a wine list, who finds the interaction with a sommelier slightly intimidating, and who would default to pointing at a familiar grape variety is better served by a tool that asks about food, flavour preferences and budget and returns a considered recommendation than by the absence of guidance.
The third is data. AI wine tools generate purchasing data — which recommendations were accepted, which wines were reordered, which pairings generated the most additional sales — that human sommeliers do not systematically produce. That data, fed back into list development and purchasing decisions, has commercial value that several operators describe as the tool's most surprising benefit.
What AI Cannot Do
The sommelier's role in a serious restaurant is not primarily informational. It is relational — the building of a rapport with a table over the course of a meal, the ability to read a shift in mood or appetite and respond, the genuine enthusiasm for a specific producer or vintage that is communicable in a way that a recommendation engine is not. It is also editorial: a sommelier who has built the list has a relationship with the wines on it that gives their recommendations an authority and a specificity that no algorithm can replicate.
"The day I find a wine tool that can tell me why this particular bottle from this particular domaine matters, and do it with the conviction that makes a guest feel the decision is real — that's the day I'll worry," says one head sommelier at a London restaurant who declined to be named. "Until then, I'll use it for the tables I can't reach and spend my time on the ones I can."
That framing — AI for accessibility, humans for depth — is where the most thoughtful operators currently position the technology. The question of whether it will eventually reach further up the quality stack is one the wine technology developers are quietly confident about and the sommelier profession is, for now, watching with guarded interest.