Guided tours, bookings, pricing: when AI handles all phone reception
Every day, an average-sized museum receives between 150 and 300 phone calls. The vast majority concern three subjects: opening hours, pricing and bookings. A human receptionist can handle only one call at a time, works eight hours a day five days a week, and costs between 1,500 and 2,500 dollars per month. During closing hours, weekends and holiday periods, the phone rings unanswered. Every missed call is a potential visitor lost, a booking that will never happen, a tourist who will choose another museum. AI ARTEDUSA offers a radical alternative: an autonomous voice AI agent, specialised for cultural institutions, available around the clock, capable of handling five calls simultaneously and speaking fifteen languages. This article walks you through the complete unfolding of a phone call managed by this agent, from the first ring to automatic post-processing.
By Artedusa
••7 min read01Fifteen tools serving museum reception
AI ARTEDUSA provides each museum with a set of fifteen specialised tools. Eight base tools are common to all cultural institutions: natural conversation enabling fluid and contextualised dialogue; RAG document search querying the institution's own knowledge base; record updating that automatically collects visitor information; reminder creation for staff or visitors; transfer to a human agent when the situation requires it; language switching mid-call; speech speed adjustment between 0.75x and 1.25x; and clean call ending that triggers post-processing.
Seven additional tools are dedicated to the museum context. The book_visit tool reserves a visit specifying date, time, number of visitors and visit type, whether individual, group or school, generating a confirmation number in the format MUS-XXXXXXXX. The book_guided_tour tool schedules a guided tour taking into account date, time, desired language and group size, with a maximum of twenty-five people and a standard duration of sixty minutes, then issues a confirmation in the format MUS-TXXXXXX. The get_ticket_pricing tool provides admission prices drawing either on the museum's configuration or the RAG knowledge base. The get_exhibition_info tool informs the visitor about current and upcoming exhibitions. The get_museum_info tool communicates opening hours, address, transport options and accessibility conditions. The book_event tool reserves a workshop, conference or special event and generates a confirmation in the format MUS-EXXXXXX. The check_membership tool provides information on subscriptions and annual passes.
02A complete call, from first ring to goodbye
To understand concretely what these tools enable, let us follow the journey of a fictional visitor, Marie Dupont, who calls a museum on a Tuesday morning. Marie wishes to visit on Saturday with two adults and one child. She picks up her phone and dials the museum's usual number. Nothing has changed on her end: same number, same ringtone. But on the museum's side, the call is handled by the AI agent.
The agent introduces itself using the museum's name, never the name Artedusa, because the solution operates as a white-label product. It offers available time slots for Saturday: 10 AM, 2 PM and 4 PM. Marie chooses 2 PM. The get_ticket_pricing tool activates and provides the rate: 12 euros per adult and 8 euros for the child, totalling 32 euros. The agent then asks whether Marie would like a guided tour. The supplement is 5 euros per person.
Marie accepts and specifies she would like the tour in French. The book_guided_tour tool reserves the guided tour for three people at 2 PM, in French, for a duration of sixty minutes. The total rises to 47 euros. The agent communicates the confirmation number MUS-T3F8A2B. The book_visit tool simultaneously registers the overall reservation and generates the confirmation number MUS-3A7B9C12. Finally, the record updating tool captures the visitor's name. The call ends with a summary confirmation and a warm goodbye.
03Data collected without human effort
During this call lasting just a few minutes, the agent collected and automatically structured a set of data in the client record, known in the voice AI world as a claim. The customer_name field contains the text Marie Dupont. The visit_date field records Saturday's date. The visit_time field sets the hour at 2 PM. The visitor_count field indicates three people. The visit_type field is recorded as individual. The confirmation_number field stores MUS-3A7B9C12. The guided_tour field contains a complete object with the date, time, French language, group size of three, duration of sixty minutes and the confirmation MUS-T3F8A2B. The booking_confirmed field is set to true.
No museum employee needed to manually enter this information. No paper form was filled in. No data was lost between the moment of the call and its registration in the system.
04Intelligence triggered after hanging up
When Marie hangs up, the agent's work does not stop. Automatic post-processing generates three distinct analyses. The first is a detailed conversation summary: a long-form paragraph and a short one-sentence summary, for example "Visit plus guided tour booking for three people, Saturday 2 PM, total amount 47 euros, satisfied visitor." The second analysis is a satisfaction score evaluated across five levels: high, partial, low, terrible or unknown. In Marie's case, the score is high. The third analysis determines the next action among six categories: case closed, commercial offer, client will send information, high priority, new contract proposal or requires expertise. For this call, the action is case closed: everything is settled, no human intervention is necessary.
The museum director finds all this information in the dashboard: the caller's phone number, call date and time, duration, complete transcription with speaker identification, the automatically filled client record, any reminders, the satisfaction score, follow-up actions and the cost of the call.
05Connection with the existing booking system
An essential point for museums already equipped with a booking system: the book_visit tool can connect directly to existing software. The agent automatically sends a request containing the date, time and number of visitors to the museum's external system. It receives the official confirmation number from the external system in return. If the external system does not respond, a fallback mechanism generates a local confirmation number. Configuration is done simply in the dashboard by entering the system address and access key. The museum does not need to replace its existing tools: AI ARTEDUSA integrates with them.
06What this changes in daily operations
An average museum receiving 150 calls per day, with an average duration of three minutes per call, generates 4,500 calls per month, or 13,500 minutes. The monthly cost with AI ARTEDUSA amounts to approximately 200 to 400 dollars, compared to 1,500 to 2,500 dollars for a human receptionist. The monthly saving reaches approximately 1,700 dollars. But the purely financial calculation captures only part of the value. A human receptionist speaks one or two languages; the agent speaks fifteen. A receptionist handles one call at a time; the agent manages five simultaneously. A receptionist is absent evenings, weekends and during holidays; the agent is available without interruption.
Training a receptionist takes weeks. Training the AI agent takes five minutes: you upload your PDF brochure, your exhibition catalogue or your pricing grid, and the agent immediately integrates them into its knowledge base through the RAG system. Documents are split into intelligent segments of intelligent segments, transformed into mathematical vectors by the embedding model, and stored in a vector database dedicated to your institution. During each call, the agent queries this database and retrieves the five most relevant segments, with a minimum relevance score of 0.8 out of 1.0.
07AI that understands art
A museum's phone reception is not a generic call centre. It demands an understanding of exhibitions, cultural practices and the specific expectations of visitors, whether a school group, a collector or a foreign tourist. AI ARTEDUSA was designed exclusively for this context. Its tools are those of the museum. Its vocabulary is that of the museum. Its mission is to allow staff to focus on what cannot be automated: in-person welcome, cultural mediation and the human relationship with visitors on site. The phone, meanwhile, is in good hands.
Discover ai.artedusa: https://ai.artedusa.com/
AI that understands art
Discover our AI agents built for museums, galleries and cultural institutions. Collection analysis, intelligent curation, personalised recommendations.
Discover ai.artedusa