The AI agent in times of crisis: when telephone reception must absorb the unpredictable
On 15 April 2019, the cathedral of Notre-Dame de Paris catches fire. Within hours, the monument's switchboard, those of neighbouring museums and the capital's tourist offices are overwhelmed by thousands of calls. Visitors want to know whether their booking is maintained. Foreign tourists, en route to Paris, seek information about the state of the district. Journalists call for confirmations. Residents worry. A single receptionist cannot absorb this deluge. Even five receptionists would not suffice. This scenario, though dramatic, is not exceptional. A flood forcing a museum's closure, a transport strike making access to an institution impossible, a pandemic disrupting schedules and capacity limits, a security incident requiring evacuation: cultural institutions face crises with a regularity that their heritage mission renders particularly sensitive. The question is not whether a crisis will occur. The question is whether your telephone reception will be capable of responding.
By Artedusa
••6 min readA human telephone switchboard, in times of crisis, reveals its structural limitations. The receptionist, themselves affected by the situation, must manage a call volume that can be ten times the normal level. They repeat the same information to each caller. They can handle only one call at a time, and while speaking with an anxious visitor, dozens of other calls are lost. They tire, their voice betrays stress, their responses become less precise as the hours pass. When evening comes, they leave their post, and calls that continue arriving go unanswered. The following day, the situation is not necessarily resolved, and the same cycle begins again.
The voice AI agent transforms this dynamic. Its capacity to handle up to five calls simultaneously is a first advantage that, in normal times, is appreciable but not critical. In times of crisis, this simultaneity becomes a lifeline. When call volume increases fivefold, the agent absorbs the load without quality degradation. It does not tire. It does not stress. It delivers the same clear and precise response to the first call as to the five hundredth. Its permanent availability, 24 hours a day, guarantees that night and weekend calls are handled with the same rigour as daytime ones.
But the true value of the AI agent in times of crisis lies in its capacity for rapid adaptation. When a crisis occurs, the first action is to update the agent's knowledge base with crisis information. This operation takes five minutes. You write a document containing essential information: the museum is closed until further notice, bookings are cancelled or postponed, here are the refund procedures, here is the number to contact for urgent cases. You import this document into the dashboard, and from that moment, the agent integrates this information into all its conversations. Every caller receives the same information, accurate, complete and up to date.
This uniformity of information is a considerable advantage in times of crisis. With a human switchboard, each receptionist may formulate information differently, omit a detail, communicate outdated information because they were not informed of an update. The AI agent has a single source of truth: the knowledge base. When this source is updated, all conversations instantly reflect the update. There is no gap between the moment a decision is made and the moment it is communicated to callers.
Consider a concrete scenario. A museum located in a flood zone is forced to close indefinitely following exceptional rainfall. The director writes a crisis document in fifteen minutes: temporary closure, estimated reopening date, booking cancellation and refund procedure, telephone number for a manager reachable for urgent cases, information about the state of collections to reassure the public. This document is imported into the knowledge base. Immediately, the agent begins answering calls communicating this information. A visitor who booked for next Saturday calls: the agent informs them of the closure, explains the refund procedure and offers to call back once reopening is confirmed. A journalist calls: the agent communicates public information about the situation. A school group due to visit the following week calls: the agent proposes postponing the visit and creates a reminder to recontact the group once the reopening date is known.
Meanwhile, the director and their team can focus on operational crisis management: assessing damage, protecting collections, coordinating with emergency services, communicating with oversight bodies. They are not consumed by a phone that rings relentlessly. The agent absorbs telephone pressure, and every call is documented in the dashboard. In the evening, the director can consult the summary of all calls from the day, identify the public's most frequent concerns and adjust their communication accordingly.
The multilingual dimension takes on particular importance in times of crisis. A German tourist calling from their hotel to learn whether the museum is open following an incident announcement does not necessarily speak French or English. With a human switchboard, this tourist encounters a language barrier at the worst possible moment. With the AI agent, they receive a clear response in their language, whether German, Mandarin, Japanese or any of the 15 available languages. In times of crisis, comprehension is paramount. A visitor who does not understand the information communicated is a visitor who risks going to the site, being disappointed, being in danger, or spreading incorrect information.
The AI agent's post-call intelligence becomes a crisis management tool. Automatic summaries of each call allow identification of the most frequent questions, emerging concerns and circulating rumours. If 30 per cent of calls ask whether collections are damaged, that signals the need to publish a statement about the works' condition. If English-language calls increase by 200 per cent, that signals the crisis has reached international media and English-language communication is necessary. If the satisfaction score drops despite correct responses, that signals the tone or content of responses should be adjusted to better address public anxiety.
The end of the crisis also requires controlled telephone management. The reopening announcement triggers a surge of calls: visitors wanting to reschedule, groups wanting to confirm new dates, media wanting to cover the reopening. The agent manages this second call peak with the same efficiency as the first. You update the knowledge base with new hours, new access conditions, any restrictions, and the agent communicates this information to every caller.
Crises do not give warning. They strike on a Friday evening, a public holiday, during the director's vacation. The AI agent, however, does not take holidays. It does not sleep. It does not panic. It responds, it informs, it documents, and it does so with the same precision in the middle of the night as in the middle of the afternoon. The question for your institution is not whether you will need this capability. The question is whether you will have it when you need it.
A prepared cultural institution is one that already has in its processes a protocol for updating the knowledge base in case of crisis. A template document, ready to be adapted and imported in minutes. An identified person responsible for performing this update. A tested procedure. The AI agent is a tool. Like any tool, its value in a crisis depends on the preparation of those who use it.
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