Fifteen languages, one agent: international tourist welcome reimagined

01The multilingual challenge for cultural institutions

Museums, galleries and cultural institutions welcome an international public by nature. The Louvre receives visitors from more than one hundred and eighty nationalities. Mid-sized museums in tourist cities see their international attendance rise every year. But telephone reception remains overwhelmingly monolingual, or bilingual at best. A receptionist who speaks French and English covers a significant share of calls. But the Spanish tourist, the Chinese visitor, the German school group, the Korean family or the Turkish couple all face a language barrier that degrades their experience and often discourages a visit.

Recruiting multilingual staff is an expensive and incomplete solution. Finding a receptionist who speaks four or five languages fluently is a rare achievement. Training them in the specifics of the institution adds further complexity. And even with such a profile, they will only be available eight hours a day, five days a week, and can handle only one call at a time.

02Fifteen languages available at all times

AI ARTEDUSA supports fifteen languages in real time. The full list covers French for France, English for the United States, Spanish for Spain, German for Germany, Italian for Italy, Portuguese for Brazil, Dutch for the Netherlands, Japanese for Japan, Chinese for China, Korean for Korea, Arabic for Saudi Arabia, Russian for Russia, Hindi for India, Polish for Poland and Turkish for Turkey.

These fifteen languages cover virtually all international tourist flows to European cultural institutions. A Parisian museum, a Florentine museum, a Madrid museum or an Amsterdam museum will find that the vast majority of its international visitors speak one of these fifteen languages.

03Mid-call language switching

The most remarkable capability is not simply the support for fifteen languages. It is the ability to switch from one language to another mid-conversation, without hanging up, without losing information, without any perceptible interruption. The agent automatically detects the visitor's language. When the visitor explicitly asks to change language, the switch is instantaneous.

The technical mechanism works as follows. The visitor's voice is transcribed to text in real time by our voice transcription engine, with latency under five hundred milliseconds. The text is analysed by our artificial intelligence, which identifies the language-change request. The agent formulates its response in the new language. The text is converted to natural speech by our speech synthesis engine, which has a dedicated voice for each language. The visitor hears the answer in their language, spoken with the right intonation, the right rhythm, the right pronunciation.

Each language has its own natural voice. Japanese does not sound like French spoken with an accent. Arabic does not carry the intonation of English. The our speech synthesis engine speech synthesis system, which is multilingual by design, produces an authentic voice for each of the fifteen languages.

04Conversational context is preserved

When the visitor switches from French to Japanese mid-call, the agent does not start from scratch. The conversational context, which includes everything said previously, is fully preserved. The agent remembers that the visitor mentioned wanting to come on Saturday with two adults and a child. It remembers the visitor expressed interest in the temporary exhibition. The conversation window retains the last twenty messages exchanged, regardless of the language in which they were spoken.

This continuity is essential to the visitor's experience. Nothing is more frustrating than having to repeat everything after a language change. With AI ARTEDUSA, the visitor simply changes the language of the dialogue and the conversation continues naturally.

05Full scenario: a Japanese tourist at a museum

A Japanese tourist calls a Parisian museum on a Sunday morning. The agent answers in French: "Bonjour, Musee des Lumieres, how may I help you?" The visitor, whose French is limited, tries to explain that they wish to visit the museum with their family. The agent detects the difficulty. The visitor says "日本語でお願いします", meaning "in Japanese, please." The agent switches instantly to Japanese.

In Japanese, the visitor explains they are in Paris for three days with their spouse and two children. They would like to visit the museum the following day, Monday. The agent consults the knowledge base and informs the visitor that the museum is closed on Mondays. It suggests Tuesday, with slots at ten, two and four. The visitor chooses ten. The agent provides the rates: twelve euros per adult, eight euros per child, totalling forty euros. The visitor asks whether a guided tour is available. The agent checks and confirms that a guided tour is available on Tuesday at ten, for an additional five euros per person, lasting sixty minutes. The visitor accepts. The agent generates a confirmation number and summarises the booking in Japanese.

After the call, the system automatically generates a conversation summary, a satisfaction score rated as "high" and classifies the next action as "case closed." The museum director sees in their dashboard that a call in Japanese resulted in a booking for four people with a guided tour. No staff member speaks Japanese. Without the AI agent, this call would have reached an answering machine and the tourist would probably have chosen another museum.

06The impact on international attendance

The language barrier is an invisible yet powerful deterrent to museum visits by international tourists. A visitor who cannot obtain information in their language before their visit hesitates more. They do not know the exact opening hours, whether they need to book, or what the prices are. Uncertainty pushes them toward a destination that is better documented online or easier to access.

An AI agent that answers in the visitor's language, twenty-four hours a day, seven days a week, removes this barrier. The visitor gets the information they need, in their language, at the time that suits them, regardless of time zone. A Chinese tourist calling from Beijing at three in the morning, Paris time, receives an immediate answer in Mandarin. This is a level of service that even the largest museums cannot offer with human staff.

07Five simultaneous calls, fifteen languages

The AI agent does not handle a single call at a time. It can process five calls simultaneously per institution, and this limit is configurable with no technical constraint. A call in French, a call in Japanese, a call in Spanish, a call in German and a call in Arabic can all take place at the same time, each with its own voice, its own conversational context, its own actions.

For a museum in a major tourist city, this capability transforms telephone reception. During peak season, when calls pour in from every time zone, every caller is attended to immediately, in their language, with no queue.

08European compliance and data protection

Multilingual processing legitimately raises privacy questions. Call data, including transcriptions in all fifteen languages, is hosted in the European Union. The system is GDPR-compliant. Data isolation between institutions is guaranteed by our secure database per-institution data isolation. Conversations are not shared between institutions and are not used to train third-party models. The system is also prepared for the EU AI Act, due to come into force in August 2026, with transparency, traceability and disclosure mechanisms for the artificial nature of the conversational intelligence.

AI ARTEDUSA was built precisely to meet this need. Find out how at ai.artedusa.com.