The context
The Jada Hotels Collection case is different from a single hotel, and the difference is not the technology: it is what happens when the same problem is multiplied and the whole tech stack has to keep up.
A single hotel has a fragmented request volume: messages come in from different channels at different times, and all of them need an instant answer, because in today's fast-moving world a guest who does not get precise information right away changes their mind. In a group like Jada Hotels the fragmentation grows: requests come not only from several channels, but from ten different properties, each with its own features, details and rates.
Picture the operational load on the front office. The exact same questions arrive across ten different pools, and every property has its own hours, its own access, its own services. At some point the problem is not just replying, it is replying in a targeted way to each guest, in their language, even when there is a check-in queue like at a supermarket till that just opened. In that scenario some requests, meaning potential bookings, get lost. And it becomes hard to study the guest, understand what they like and dislike, and offer upsell and ancillary services.
Jada Hotels runs ten properties across Rome, Tuscany and Umbria. We started on February 17, 2026 with a pilot property and, once the added value was clear, rolled out to all ten within a few months. The numbers in this case study come from real production data.
From one property to ten in six months
Let's start with the volume handled by the agents. Monthly conversations went from 57 in February to 1,445 in August, twenty-five times as many in six months.
The interesting part is not the growth itself, it is where it comes from. It is not one property getting more traffic: it is properties going live one after another and producing from day one, each with its own guest pool. Together the ten properties handled 1,785 pre-stay conversations, and the group's in-stay assistant another 2,249, for a total of 4,034 conversations. All without adding a single person to the front office.
This is what matters for anyone running a cluster or, more broadly, multiple properties: adding a property costs what the previous one did. For a group growing through acquisitions or affiliation, it is the difference between a cost that balloons and a cost that scales with the group.
Seventy-five countries
Guest communication inefficiencies are not only about knowing every property's details, they are also about speaking the guest's language. On messages and email you can translate the text; on the phone it becomes a problem.
Demand data shows the group is 44.8% international: 1,100 conversations in English against 1,534 in Italian, plus Spanish, German, French and Portuguese. Guests come from 75 countries and 781 cities. The top foreign markets are Germany (90 conversations), the United Kingdom (70) and the United States (57), followed by Spain, France, Ireland, Poland and Switzerland.
Talking to hotels, language always comes up as a translation problem. Looking at these numbers, the real problem is something else: the combination of language and time of day. With guests living in another hemisphere, often high spenders, replying quickly and well is complicated by the time zone. Covering that demand by hand would mean multilingual staff on duty around the clock across the whole cluster. That is the space for a technology that literally answers the phone and messages at night, with detailed knowledge of all ten properties. At that point ROI is beside the point: it is not a saving, it is a capability that did not exist before.
There is more. In a group like Jada Hotels, studying where guests come from becomes strategic. Hotel Relax Roma Nord has 13.6% foreign guests, Rome Ciampino Smart Hotel 40.3%, with more than half of its requests arriving out of hours. Two almost opposite demand profiles, inside the same group, a few kilometres apart. Until you measure them per property you keep treating them as the same thing, with the same pricing setup and the same way of presenting the offer.
Does it work?
3,356 of 4,034 conversations closed with no human involvement, 83.2%. Of 28,161 total messages, 12,486 are guest questions and 10,825 assistant replies.
The first reply arrives in a median of 8.5 seconds, and within 14.7 seconds in 90% of cases. That speed is not a matter of style, it is commercial: in those 9 to 14 seconds the assistant prepares a quote on the rooms that match exactly what the guest had in mind, and shows them, explains them, presents them in a way designed to make the guest like what they see. That is conversion, and conversion guests enjoy, because our system is designed to be warm and hospitable, to suggest and propose, selling without selling. It gets even more interesting when the system sells based on demand pressure: when the property is filling up, alternative quotes become a tool to maximise occupancy.
Beyond the content there is time. A guest who writes and gets an answer in under ten seconds has no time to open another tab and write to another property. First reply time is not a service metric, it is a conversion metric.
Back to the numbers: for bookings with check-in in 2026, the assistant talked to the guest in 2,042 of 6,217 cases, 32.8%. That means every result in this document (autonomy, quotes, problems caught, markets identified) comes from one third of booked guests. The other two thirds typically come from OTAs without usable contact details, so they never reach the assistant.
That is the real measure of today's coverage, and also the biggest lever available: it needs no new technology, only connecting the contact details of indirect bookings.
From pre-stay to stay
Pre-stay is handled property by property, with a dedicated assistant for each. The stay is covered by a single assistant across the whole group.
Looking at autonomy, pre-stay properties close between 91.6% and 96.5% of conversations with no human involvement, while the in-stay assistant drops to 74%. At first glance it looks like a step back. It is actually exactly what should happen.
The reason for the hand-off, passing the conversation to the front office, lies in the type of request. Before arrival the questions are informational and repetitive (hours, access, rooms, availability), and that is where automation wins back almost all the time. During the stay the questions become operational: a delivery, a room change, a fault. Handing that to a person is not a failure, it is the solution.
And the relationship between AI agents and human staff is becoming more and more interesting. Responsiveness on housekeeping requests and in-stay problems has a direct impact on guest satisfaction and on the reviews requested at the end of the stay. Yes, those are automated too.
The guest profile CRMs make you fill in by hand
Every conversation, from pre-stay through the stay and after check-out, is analysed, and the information extracted becomes the basis for further value. Without asking the front office to fill in guest cards, when check-in already keeps them busy enough.
Across almost 3,000 analysed conversations:
- 809 contain a booking request (27.1%). That is demand expressed by an interested guest, whose contact becomes a lead. It never goes through the PMS and, without this analysis, would not exist in any database.
- 671 contain references to places and the local area (22.5%), the second largest category. Analysed with the right algorithms, this unstructured data gives real-time demand pressure on what guests want from their stay, which services they ask for, where they want to go and what they want to see. At Jada, guests ask more about what to do than about which room to book.
- 624 are service requests (20.9%): upsell candidates already validated by the guest, not marketing hypotheses.
- 398 contain a stated preference (13.3%), the basis for personalising the offer when the guest returns, again in a consistent, steady and autonomous way.
Insights like these, connected to a digital concierge, become ancillary revenue. Upsell and cross-sell not done at random, but suggested to the right guest, with a given profile and segment: an offer you already know will convert and that, through its naturalness, precision and detail, feels sold without being sold.
This is not only about growing ancillary revenue, it is about amplifying the experience. Suggesting a specific service or experience to a specific guest, at the right time and in the right tone, is hospitality becoming precise, rare, distinctive. In other words, luxury.
Problems that surface while the guest is still on site
Of those conversations, fortunately, around 17% are used to report problems. Why fortunately? Because going through a digital concierge makes it possible to analyse, manage and act on problems proactively and immediately. The inconvenience becomes a chance to delight.
510 conversations contain a problem reported by the guest. They are not reviews and they are not complaints sent afterwards: they are problems raised in chat during the stay, while the guest was still on site and the issue could still be fixed.
That window usually does not exist. In most cases the problem goes straight to Booking or Google as a negative review. A hotel's normal cycle is finding out after the fact, three days after check-out, with the review already online for everyone to see. Here the problem surfaces while the guest is writing to ask about something else, and that is the only situation in which stepping in prevents the negative review instead of having to manage it.
What that is worth in euros is hard to quantify precisely. What I can say is that a negative review hurts online reputation, which in turn directly affects how much you can charge for a room. And 510 chances to step in over six months across ten properties is a number no manual process produces.
Even complaining in chat becomes data for us. Our systems track guest sentiment in real time: they alert the front office when it turns negative and stop the automatic review request from going out at the end of the stay.
What a hotel group can take from this case
Cost does not scale with properties. Adding a property costs what the previous one did. For a group growing through acquisitions or affiliation, it is the difference between a cost that balloons and a cost that scales with the group.
International demand is a capability, not a saving. 44.8% of guests are foreign, from 75 countries. Covering that by hand would mean multilingual staff on duty around the clock across the whole cluster.
Properties in the same group have opposite profiles. Hotel Relax at 13.6% foreign guests, Rome Ciampino at 40.3%. Until you measure them per property you keep treating them the same way, with the same pricing setup.
Conversations are the data. 809 booking requests, 671 local-area references, 624 service requests and 398 stated preferences are the basis on which upsell and cross-sell stop being random.
Good automation does not close everything on its own. Pre-stay closes over 90% autonomously because the questions are repetitive, the stay 74% because the questions are operational and handing them to a person is the right answer. They are two different labour economies, and they need two different metrics.
If this is interesting to you, let's talk, it interests me too, maybe over a coffee. To see how it applies to your group, start with the digital concierge and AI for hotels.
