What Do Customers Really Ask an AI Sommelier? We Analysed 687 Conversations to Find Out

1. It’s not just about the wine. Customers expect an AI Sommelier to know everything about your store. Prices. Pairings. Recommendations. Gifts. Opening hours. Delivery times. Customers ask the AI everything about your business.

2. Customers want specific pairing recommendations.Roasted lamb shank. Barbecued salmon. Freshly shucked oysters. The meal is already planned – the wine, beer or spirit has to match the meal.

3. Customers need nuanced recommendations.Customers expect the AI to ‘get what I mean’ – using qualitative expressions such as ‘sweet, but not too sweet.’

4. Customers expect personalised gift suggestions.Not just a generic suggestion or a gift for a specific price, but a gift suitable for a specific person or occasion – an anniversary, a 21st birthday, a friend who loves cigars.

5. Customers expect to ask and be answered in their own language. Across our dataset, customers automatically asked questions in their own language – sometimes even switching language mid-question.

1. Nearly 1 in 3 Questions are Operational. 28% of conversations were about store operations — delivery, orders, opening hours, refunds, discount codes…

• I placed my order 5 days ago and it’s still not ready
• Can wine be delivered to Hotel Furama at 12am?
• Am I gonna get my refund back?
• What’s the $5 discount code?
• Is click and collect available?
• Can I arrange delivery for Wednesday specifically?

2. Food Pairing Requirements are Very Specific

• I’m cooking lamb shanks, what red wine under €15 can I get?
• Wine for a BBQ
• What bourbon pairs well with chicken?
• Ich weiß nur Beef Stroganoff. Budget so um die €15′ (German: I only know it’s beef stroganoff, budget around €15)

3. Customers Want Personalised Gift Suggestions

• Whisky for my friend loves cigars
• Whisky as a gift for a good friend I haven’t seen in a long time
• I am writing from outside of Ireland hoping you can help me send a special Christmas gift to a friend who lives there
• Preciso também presentar uma amiga, pode me surpreender com algum vinho em torno de 50
• Chinese New Year Gift for around $200

4. Customers Expect Expert and Personalized Guidance. Customers want more than just generic recommendations. Some of the most interesting conversations were about style and flavour preferences — nuanced questions that keyword searches and menus can’t respond to:

• Red wine dry, smooth, tannins, bold, dark fruits, chocolate
• A chardonnay for under $30 that is not buttery
• I would like a whiskey that can be enjoyed neat. Preference for smoky
• What Spanish wine for an experienced drinker with luxurious tastes for €18–€25?
• I like Guigal’s Côte du Rhône, Rhinforzo Primitivo and Tuscan reds can you recommend something similar under €20?•
• Can you recommend a wine that is sweet but not too sweet?
• I love Tasmanian Pinot and Eden Valley Cabernet have you got something that matches that mix?

5. Customers Expect to Communicate in Their Own Language. Across our dataset, questions were asked in English, German, French, Dutch, Italian Spanish and Portuguese — without merchants configuring anything. Customers simply typed their questions in their own language.

• Tienes tequila de México? (Spanish)
• Preciso presentar uma amiga (Portuguese)
• Ich weiß nur Beef Stroganoff (German)
Suggest a wine from Europe that pairs well with roasted salmon (English)

The questions demonstrate this – when customers see an AI Sommelier, they jump on it, asking it all sorts of questions. Customers treat the AI Sommelier the same way they would treat a member of staff, not as a specialist tool. They ask whatever they need to know, regardless of what the AI Sommelier was designed for. That means the AI agent needs to be as well informed as the store staff. It needs to be able to provide pairing advice, know the store’s live inventory as well as all the information about delivery, discount codes, opening hours in order to meet customers’ expectations.

In-store, customers can ask the staff all the questions they need answers. The staff are trained, informed, patient and available. Online, customers are asking the same questions, but no-one is answering. Instead, customers are faced with filters and questions, but no-one is answering. Instead, customers are faced with filters and drop-down menus – no guidance and no personalized advice. An AI Sommelier trained on your inventory, company operations and policies and capable ofcommunicating in natural language can answer all of the questions customers are asking online.

What does this mean for online wine and spirits retailers. In their conversations with the AI Sommelier, customers are sharing context that In their conversations with the AI Sommelier, customers are sharing context that they would never put into a search box. It’s the same depth of information-sharing that an in-store wine expert needs to know when helping an in-store customer — the relationship, the occasion, the person receiving the gift. The conversations show that customers expect the AI Sommelier to match that level of conversation.

Expert-level customers expect to be able to use the kind of language they would use in-store when chatting with a sommelier. These customers know exactly what they want and they use the kind of comparative and evocative language than a search engine can’t cope with. Natural language – the kind that AI agents are trained in – is the only interface that works for these kinds of online queries.

Without prompting, customers typed their questions in their own language, and were answered in the same language. Customers expect that nothing gets lost in translation. For retailers serving international customers — whether through export shipping or a multilingual local customer base — a multilingual service is essential. It’s the difference between making a sale and losing a customer who couldn’t get answers they understood.

• Your AI needs to know your store and your inventory in detail. Nearly a third of questions are operational. If the AI can’t answer them, it fails the customer

• Conversational search is a powerful addition to online service. Food pairing, style preference, gifting context and nuanced recommendations — these are the questions that drive real purchase decisions, and none of them work in a keyword search bar

• Multilingual response ability is a basic requirement, not a premium add-on. Customers will ask in whatever language they’re comfortable in. If the AI can’t respond, the sale is lost.

• Gift occasions are high-value opportunities. Customers buying gifts are often spending more, and they’re sharing the context that makes a great recommendation possible. This is where a good AI shows its value.

Dataset compiled from 687 real customer-AI Sommelier conversations across three continents, Jan – Mar 2026