It's 11:47 pm. An Italian couple has just landed at Roissy two hours late. They have booked an Airbnb apartment in the 11th arrondissement, they are tired, they don't speak a word of French, and the lockbox code you sent them via Airbnb message three days ago doesn't work. They call you. You're asleep. They call back. They leave a message. They open an Airbnb dispute. The next morning, you've lost 1.2 stars on your host profile and 280 euros in partial refund. This scene plays out tens of thousands of times every night in all European capitals.
The Airbnb concierge business has exploded in five years: according to data from the Observatory of Furnished Tourist Rental, France now has more than 4,200 professional concierge companies each managing between 5 and 400 properties. The problem is no longer finding properties to manage, it's absorbing the volume of guest calls without burning out the teams. An apartment generates an average of 3.2 calls per stay. For an 80-property concierge with an occupancy rate of 75%, that's 70 to 110 calls per day, of which 22% are outside regular office hours.
Airbnb concierge in 2026: call volume and multilingual challenge
The business has changed in nature. What was ten years ago a marginal service reserved for a few Parisian pieds-à-terre has become a structured industry with its standards, processes and bottlenecks. The phone remains channel #1 — not Airbnb chat, not email, the phone — because a traveller lost on an unknown street at midnight is not going to wait for a written answer.
The real breakdown of inbound calls
On a sample of 18 concierges managing between 25 and 180 properties, the typology of inbound calls breaks down as follows:
- Check-in and access (38%): lockbox code, building instructions, elevator, floor, Wi-Fi
- Practical requests (24%): heating, washing machine, TV, air conditioning, extra cleaning
- Local recommendations (12%): restaurants, metro, supermarket open on Sunday, on-call pharmacy
- Technical emergencies (9%): water leak, heating breakdown in winter, jammed lock, no more hot water
- Stay changes (8%): late check-out, early arrival, extension
- Complaints and disputes (5%): noisy neighbours, insufficient cleanliness, missing equipment
- Miscellaneous calls (4%): cold calling, wrong numbers, late travellers
Out of these seven categories, six are scriptable and can be automated. Only real technical emergencies and complex disputes require immediate human intervention — and even then, not always in the middle of the night.
The multilingual challenge specific to short-term rentals
Airbnb is by nature an international product. Travellers are foreign in 64% of cases for Paris, 71% for Barcelona, 58% for Amsterdam, 49% for Lisbon. A serious concierge must be able to welcome a Japanese traveller at 11 pm, a Brazilian couple at 1 am and a Polish family at 4 am. The reality on the ground: most concierges operate in French and decent English, plus two or three approximate languages. Anything outside this perimeter becomes an operational nightmare.
Calls handled by AI: check-in, codes, instructions, emergencies
The voice agent is not an improved answering machine, it's an assistant that conducts a real conversation, identifies the traveller, understands their request and executes concrete actions — sending an SMS with a code, remotely opening a connected door, creating a ticket in your PMS or escalating to on-call.
End-to-end autonomous check-in
When a traveller calls for check-in, the agent follows a precise protocol. It first identifies the booking by the traveller's name or the Airbnb reference (10-character ID). It verifies the arrival date in the calendar of the property concerned. It communicates the lockbox code or, if you use a Nuki, Igloohome or TTLock smart lock, it triggers the opening for the planned time window. It then gives the practical instructions: building digicode, floor, door, Wi-Fi code, heating operation. If the traveller expresses doubt ("I can't find the building"), the agent reconfirms the exact address and can send an SMS with a precise Google Maps link.
Forgotten codes and information retrieval
One of the most frequent calls is "I've lost the Wi-Fi code" or "the door code no longer works". The agent retrieves the information from the property sheet, communicates it verbally and sends it in parallel by SMS so the traveller can keep it. If the problem persists (recently changed code, traveller in the wrong building), it escalates to the human team with all the context already recorded — no need to start over from scratch.
Comfort and operating requests
"How do you turn on the heating?" "The TV doesn't work" "Where is the washing machine?" These questions represent a quarter of the calls and are perfectly scriptable. The agent knows each property in your portfolio with its specificities: Netatmo thermostat in the studio on rue de Rivoli, manual boiler in the Belleville apartment, reversible air conditioning in Cannes. It guides step by step, and if the procedure fails, it schedules a technician intervention for the next morning without bothering your on-call.
40-language multilingual: the international Airbnb advantage
This is where voice AI takes a decisive advantage over any human team, even one that is very well staffed. The agent automatically detects the language within the first seconds — it hears "hello", "hola", "ciao", "guten Tag", "konnichiwa" and switches without asking for confirmation. No transfer to a colleague who speaks Italian, no "can you repeat in English please", no shame in admitting you don't understand.
The 40 languages natively covered
The agent handles without quality degradation: French, English, Spanish, Italian, German, Dutch, Portuguese (Portugal and Brazil), Mandarin, Japanese, Korean, Russian, Arabic, Polish, Swedish, Norwegian, Danish, Finnish, Czech, Hungarian, Romanian, Bulgarian, Greek, Turkish, Hebrew, Hindi, Indonesian, Thai, Vietnamese, Ukrainian, Croatian, Serbian, Slovak, Slovenian, Lithuanian, Latvian, Estonian, Catalan, Basque, Galician and Icelandic. The coverage corresponds exactly to the top 40 nationalities travelling via Airbnb in Western Europe.
Why multilingual changes the customer relationship
A traveller welcomed in their native language rates their stay 0.4 stars higher on average on Airbnb, according to behavioural studies conducted in 2024 on 12,000 stays. For a concierge whose revenue depends on Airbnb ranking (Superhost, top placements in search results), 0.4 stars cumulative over 12 months represents a revenue variation of 15 to 22% per property. Multilingual is not a comfort, it's a measurable revenue lever.
The other benefit is less visible but just as real: the drastic reduction in misunderstandings that escalate into Airbnb disputes. Many complaints raised via the platform simply come from a nocturnal misunderstanding between a traveller speaking approximate English and an exhausted on-call team. When everyone expresses themselves in their language, these frictions disappear.
Human escalation for real emergencies: leak, breakdown, serious problem
The goal is not to replace the human team, it's to protect it for situations where its added value is real. An active water leak at 3 am requires a quick decision and plumber-traveller coordination that only a qualified human can take. The AI agent recognises this type of situation and triggers the escalation in seconds.
The 3-question emergency qualification protocol
As soon as a traveller mentions a technical problem, the agent chains three questions calibrated to assess the criticality:
- Material criticality: is it an active water leak, a gas leak, a general power outage, a totally jammed lock, ongoing damage?
- Risk to people or neighbours: is there a risk of spreading to the apartment below, a risk to the traveller's health, a call from the neighbours?
- Possibility of deferring: can the situation wait until tomorrow morning without significant degradation, or do we need to intervene within two hours?
Depending on the answers, two paths are triggered. Either the agent handles the situation itself (troubleshooting procedure, morning technician appointment booking, sending a temporary solution by SMS). Or it transfers to the human on-call with a complete summary already written: address, traveller, nature of the problem, assessed criticality, actions already attempted.
Typical escalations that justify a human call
- Active water leak with impossibility to cut the general supply
- Gas smell or suspicion of gas leak
- General power outage that cannot be resolved by the circuit breaker
- Traveller locked out of the property with impossibility to use the smart lock
- Neighbour conflict with probable police intervention
- Medical problem of the traveller (fall, malaise)
- Airbnb dispute involving a refund above a threshold defined by you
The rest — faulty heating one October evening, broken washing machine, TV that no longer picks up, changed Wi-Fi code — is handled by AI with a troubleshooting procedure and a technician appointment scheduled the following day. You no longer wake up for that.
Concrete case: 80-property concierge in Paris, -78% night calls
To concretely illustrate what the voice agent changes in the operational daily life of an Airbnb concierge, here is the detailed feedback from a Parisian structure — anonymised at the request of its director — that manages 80 apartments spread over the 9th, 10th, 11th, 18th and 20th arrondissements, with an average occupancy rate of 78% over the last twelve months.
The situation before the voice agent (October 2025)
- Average volume: 92 inbound calls per day (peak season July-September: 138/day)
- Team: 3 people during the day (8 am-8 pm) + 1 rotating night on-call
- Overnight on-call cost: €4,100 / month (3 people in rotation, night premium included)
- Missed calls: 14% over the month, mainly between 11 pm and 6 am
- Airbnb disputes opened for communication problems: 11 over the quarter
- Average Airbnb rating on 80 properties: 4.71 / 5
- Foreign travellers: 67%, of which 23% speaking neither correct French nor English
The situation after 6 months of voice agent (April 2026)
- Identical average volume: 94 calls per day (stable occupancy)
- Calls fully handled by the agent: 73 / day (78%)
- Calls escalated to the human team: 21 / day (22%)
- Overnight on-call calls received by the manager: 0.9 / night on average (vs 4.7 before)
- Overnight on-call team reduced to 1 backup person only
- Monthly on-call saving: -€2,800 / month
- Missed calls: 0.4% over the month
- Airbnb disputes for communication problems: 2 over the quarter (-82%)
- Average Airbnb rating: 4.86 / 5 (+0.15 stars)
- Superhost rate maintained on 100% of eligible properties
"We manage 80 apartments in Paris and the night on-call had become our biggest management problem — turnover, fatigue, mistakes. Since we deployed the voice agent last November, my phone rings on average once a night, compared to five before. The gain is not just financial: we have found a collective that no longer collapses with fatigue on Monday morning. And Italian, Spanish, German travellers — they have become our best reviews."
— Manager of a Parisian Airbnb concierge, 80 properties, May 2026
The detailed breakdown of remaining escalations
On the 22% of calls that still reach the human team, here is the breakdown:
- 9% real technical emergencies (leaks, critical breakdowns)
- 6% disputes requiring a managerial decision (refund, commercial gesture, neighbourhood management)
- 4% atypical cases not scripted (luggage lost at the neighbour's, alleged theft, medical situation)
- 3% specific requests requiring a PMS or accounting verification
None of these cases could be handled by a rigid human script either — these are situations that require judgement and decision authority. This is precisely the value you want to preserve in your human team, freed from the 78% of routine calls.
Integration with industry tools
The agent integrates with the main PMS on the market: Smoobu, Lodgify, Hostaway, Beds24, Avantio. Traveller sheets, occupancy calendars and access codes are synchronised in real time. On the smart lock side, native integrations cover Nuki, Igloohome, TTLock, August and Yale Connect. For cleaning and technical intervention planning, the agent can write to Doinn, Properly or Breezeway. For concierges working without PMS, the agent operates on a Google Sheets base that you keep up to date, with a daily automated update from your Airbnb extranet.
Frequently asked questions from hosts and concierges
Can voice AI really handle an Airbnb check-in without human intervention?
Yes, in 92% of cases according to data from deployment on 1,200 active Airbnb properties. The agent identifies the traveller via their booking name or the Airbnb reference, communicates the lockbox code or controls the opening of a smart lock, gives the Wi-Fi instructions, floor, building code and answers location questions. For the 8% of atypical cases (wrong address, lost traveller, recently changed code), the agent escalates to the human concierge with all the context already recorded — without starting from scratch.
In how many languages can the agent greet Airbnb travellers?
40 languages natively, including English, Spanish, Italian, German, Dutch, Portuguese, Mandarin, Japanese, Korean, Russian, Arabic, Polish and Swedish. The agent automatically detects the language within the first seconds of the call and switches without asking for confirmation. To go further on this subject, see Voice AI agent 40 languages.
How does the agent distinguish a real emergency from a comfort request?
The agent follows a 3-question qualification protocol: material criticality (water, gas, electricity), risk to the traveller or neighbours, possibility of deferring until the next morning. An active leak triggers an immediate SMS to the on-call manager with address, traveller and exact nature of the problem. Faulty heating at 10 pm, on the other hand, triggers a phone troubleshooting procedure, then a technician appointment scheduled for the next morning — without waking anyone up.
How many calls can an 80-property concierge delegate to AI?
On Airbnb concierges managing 50 to 150 apartments, the agent handles on average 78% of inbound calls without human escalation: check-in, forgotten codes, Wi-Fi, restaurant recommendations, practical requests. The remaining 22% concern genuine technical emergencies and complex disputes, handled by the human team during the day or by on-call at night. For related approaches, see also Short-term rental management AI and Automated property management.