Our work
TheFork
BotiqueAI
TheForkLLM · Conversational searchRestaurants · Booking
Case Study

Conversational Restaurant Search

From filters to conversation: finding a restaurant by describing it in your own words.

TheFork is a restaurant reservation platform. We built a conversational assistant deployed in production: users describe what they want in natural language and the system retrieves the most relevant restaurants based on context, real-time availability and preferences.

Prod
in productionDeployed and monitored assistant
Chat
natural languageText or voice, no filter menus
Live
availabilityResults reflect real-time availability
LLM
monitoringAnswer quality tracking
Client
TheFork
Restaurant reservation
Industry
Restaurants
Online platform
Solution
Conversational assistant
Natural-language search
Input
Text or voice
“A dish, a vibe, a plan”
Status
In production
With LLM monitoring
The Challenge

Choosing a restaurant
the way people talk about it.

On a booking platform, you filter by cuisine, location and price. But that is not how people decide: they think “somewhere romantic with a view tonight”.

The challenge: turn that free-form intent into a selection of restaurants that can actually be booked.

Filters that do not capture intent
Cuisine, location and price say nothing about atmosphere, occasion or timing.
Understanding context
The same sentence implies a time, a place and preferences that must be interpreted correctly.
Only suggesting what can be booked
A good answer takes real-time availability into account.
The Solution

An assistant that guides
instead of filters.

Users talk to the assistant instead of navigating menus, and get restaurants that match their request.

①
Understanding the request

Users phrase their request freely, by text or voice: a dish, a vibe, a plan. The assistant extracts the intent and context.

LLMLangage naturelSaisie vocale
②
Finding relevant restaurants

The system retrieves the most relevant restaurants based on context, the user’s preferences and real-time availability.

Recherche contextuellePréférencesTemps réel
③
Production and LLM monitoring

The assistant is deployed in production, with LLM monitoring to keep answer quality under control over time.

ProductionMonitoring LLM
Impact

A new way
to search.

Chat
instead of filtering
Closer to how people decide
Live
availability
Bookable results
Prod
deployed
In real conditions
BotiqueAI
Custom conversational AI assistants
← Back to portfolio