Fictional testimony: one week in the life of a museum equipped with an AI agent

The Musée des Lumières is a mid-sized municipal museum located in a French city of 80,000 inhabitants. It employs fifteen people, including a full-time receptionist, Claire, who handles phone reception Tuesday to Friday, from 9 AM to 5 PM. The museum welcomes approximately 45,000 visitors per year and receives an average of 120 phone calls per day. For three weeks, the Musée des Lumières has been using a voice AI agent for its phone reception. Here is the account of its first typical week.

Monday morning, 7:30 AM. The museum is closed on Mondays. Normally, calls go to voicemail. This Monday, the AI agent picks up. A teacher calls from Strasbourg to prepare a school trip with their Year 5 class. They want to know whether the museum offers educational workshops suited to ten-year-olds, how much a visit costs for a group of twenty-eight pupils and three chaperones, and whether the museum is accessible by coach. The agent consults the museum's knowledge base and answers point by point. The school rate is 5 euros per pupil, free for chaperones. The museum offers a "Lights and Colours" workshop for Years 4-5, lasting 45 minutes, at 3 euros per pupil. The coach park is 200 metres from the entrance. The agent books the visit for Thursday 15 May at 10 AM, for 28 pupils and 3 adults, with the educational workshop. The confirmation number is communicated. The teacher hangs up satisfied. It is 7:42 AM. No museum employee has arrived yet.

Tuesday morning, 10:15 AM. Three calls arrive simultaneously. Claire is responding to a visitor at the physical reception desk. The agent handles all three calls. The first is an Italian tourist asking in English whether the museum is open on Wednesday and how much admission costs for two adults. The agent responds in English, gives the hours and prices, and offers to book. The tourist accepts. Second call: a man asks in French about the temporary exhibition's hours. The agent consults the knowledge base and provides dates, times and a brief description. Third call: a woman calls to cancel a reservation. The agent records the cancellation and confirms. All three calls are handled in under eight minutes. Claire was not interrupted once.

Wednesday, 2 PM. A group of Japanese tourists is in town. Their guide calls the museum to confirm that the group of twelve will indeed be welcomed at 3 PM as planned. The call begins in French but the guide quickly switches to English, then asks "Can the tour be in Japanese?" The agent confirms it can provide information in Japanese during the call and that the museum does not offer human-guided tours in Japanese, but that the audio guide is available in Japanese. The guide is satisfied with this clarification and confirms the group's arrival.

Thursday morning, 6:45 AM. A Chinese tourist calls from their hotel. It is 1:45 PM in Shanghai. They wish to visit the museum on Friday with their family of four. The agent responds in Chinese, proposes available slots, gives the family rate, makes the booking and communicates the confirmation number. The call lasts two minutes and forty seconds. When the museum's director, Isabelle, checks her dashboard at 9 AM, she discovers that the agent handled four calls between 6 and 8 AM, including two in foreign languages. Before the agent's installation, these calls would have gone unanswered.

Friday, 4:30 PM. A visitor calls with a complaint. They visited the museum that morning and feel the temporary exhibition did not match the description they had read. The agent listens to the complaint attentively, records the details in the visitor file, assigns a "low" satisfaction score and recommends a "high priority" action. It offers to transfer the visitor to a manager. The visitor accepts. The call is transferred to Claire, who picks up the conversation with the full context already recorded. She does not need to ask the visitor to explain everything again.

Saturday, 11 AM. The museum is at its weekly attendance peak. Claire is deployed at the physical reception. The agent handles all phone calls. Between 10 AM and 1 PM, it processes 31 calls: 18 enquiries about hours and prices, 7 visit bookings, 3 enquiries about the temporary exhibition, 2 calls in English and 1 in Spanish. Not a single call went unanswered. The overall satisfaction rate, calculated automatically from the analysis of each conversation, is 94 per cent.

Sunday, 9 PM. Isabelle reviews the weekly summary on her dashboard from home. In one week, the agent handled 347 calls. It made 43 bookings, including 8 for groups of more than ten people. It answered 29 calls outside business hours, including 11 in foreign languages. The average conversation time is 2 minutes and 48 seconds. The satisfaction rate is 91 per cent. The week's cost is 87 dollars.

Isabelle mentally compares with the previous situation. Before the AI agent, Claire handled approximately 60 calls per day, Tuesday to Friday, or 240 calls per week. The remaining 107 calls, those on Monday, in the evening, at weekends and those arriving when Claire was already on a line, were lost. The agent therefore answered 107 calls that would have been missed, of which 43 converted into bookings. At an average admission of 12 euros and an average of 2.5 visitors per booking, these 43 bookings represent approximately 1,290 euros in additional revenue. For a cost of 87 dollars.

Claire, for her part, appreciates the change. She spends less time on the phone and more time welcoming visitors in person, guiding them through the galleries, answering questions about the works. She has rediscovered the heart of her profession. The agent handles the routine calls, the hours-and-prices enquiries repeated dozens of times each day. And when a visitor needs to speak to a human being, the agent transfers them with the full conversational context.

This account is fictional. But the figures are realistic, the functionalities are real, and the situations described are those French museums experience daily. The week you have just read could be your museum's next week.

Discover ai.artedusa: https://ai.artedusa.com/