The last question had placed us once again on the other side of the door. During the previous conversations, we had gone deeper and deeper into the hotel, trying to discover where the knowledge accumulated by the operation lived and what needed to happen for that knowledge not to depend exclusively on the people who had acquired it. But the future guest could not accompany us on that journey. For them, all that knowledge would remain invisible unless it left some trace outside the property.
A hotel could know perfectly well that certain rooms were better suited to someone particularly sensitive to noise, that an arrival outside the usual hours could be coordinated under certain conditions, or that some dietary needs could be accommodated if they were known sufficiently far in advance. That knowledge could be well distributed among reception, reservations and food and beverage, form part of the team’s experience and even guide decisions consistently. None of that meant, however, that someone could discover it before booking.
I returned then to imagining someone looking for a hotel, although this time the scene was slightly different from the one with which we had begun our journey. They were no longer necessarily going through a list of properties, opening one page after another. They could ask a question much closer to the need they were actually trying to address: they were looking for a quiet hotel for a trip with their mother, who was a light sleeper; their flight would arrive late and they needed to know whether they could check in after the usual hours; and it would also be important for breakfast to accommodate a dietary restriction.
They were not simply looking for a collection of services. They were describing a situation, and that difference gave a new meaning to much of what we had been working through over so many conversations. Answering such a question required more than knowing that the property had rooms, breakfast, reception or a particular category. We needed to understand whether some specific characteristics of the hotel could reasonably relate to that person’s needs.
An experienced receptionist would probably have known how to have that conversation. They would ask what exactly “quiet” meant, know the expected arrival time, enquire about the dietary restriction and explain what could be done, under what conditions and what should be arranged in advance. They would not respond solely from a list of services; they would use their knowledge of the hotel to interpret a situation.
Before arrival, however, that same conversation could be taking place without anyone from the property participating in it.
A search engine, a travel platform or an artificial intelligence system could try to respond using whatever it found available: the hotel’s website, its policies, photographs, business profiles, reviews, publications, third-party information and any other sufficiently accessible signal. It could connect them, find consistencies and even reconstruct a reasonably accurate picture of the property. We ourselves had done so throughout much of this journey. But none of those systems could consult what remained only in the memory of the person who knew the rooms best, in the experience of whoever coordinated late arrivals or in what the restaurant team had learned after years of dealing with particular needs.
The problem was no longer simply that the hotel knew something and the guest did not. Systems capable of participating in the choice could also stand between them, and their ability to understand the property necessarily depended on what could be found outside it.
That brought me back to a temptation we had encountered before: thinking that the solution was simply to produce more information. If we wanted to be understood, it seemed reasonable to explain more things, describe more services, publish more content and make everything we knew available somewhere.
Yet a hotel could fill the Internet with information and still be difficult to understand.
We had seen something similar near the beginning of the book. Rooms, breakfast, location, Wi-Fi, parking, views, gastronomy and service could all be accurate pieces of information and still fail to explain with sufficient precision what kind of experience they created together or for whom they might be particularly valuable. Now the same problem was appearing from another direction: making internal knowledge visible could not mean indiscriminately pouring it outside.
No future guest needed to know the procedure reception used to organise an arrival outside the usual hours, the internal conversations that helped determine which rooms were appropriate for certain preferences or the way the kitchen and restaurant teams coordinated a particular adaptation. What they needed to know was much more concrete: if they arrived late, whether they could enter; if they particularly valued quiet, whether there were differences between rooms worth considering; if they had a dietary restriction, whether the hotel could accommodate it and needed to know about it in advance. The answers might depend on a considerable amount of operational knowledge, but what needed to be understandable outside could be much simpler than everything required to produce them.
I began to distinguish between explaining the operation and making what the operation had learned understandable. A policy could say that arrivals after a particular time required prior coordination. That sentence did not explain how reception organised itself internally, but it made a relevant condition understandable before someone chose. A precise description of the rooms could help someone recognise certain differences without revealing the internal criteria used to allocate them. A clear indication concerning dietary needs could explain what was possible and when it needed to be communicated without turning the breakfast page into a kitchen procedure.
What the hotel knew could thus become a signal without becoming a manual, although not every signal allowed someone to understand with the same degree of precision.
I remembered then what we had discovered when trying to decide when a signal deserved trust: a claim repeated in many places could still have a single origin; a general phrase could sound convincing and yet require too many assumptions to connect it with a particular need; an old policy might no longer describe the current operation; a detailed review could bring us very close to an experience while still remaining the experience of one person.
At that point, we had looked at the problem from the outside and asked how much we could trust what we found. Now we could ask the inverse question: if the hotel knew something important about itself, what trace did it need to leave so that someone could understand it without unnecessarily filling in what was missing?
The answer was unlikely to lie in finding a more attractive phrase. Personalised service might express an intention, but it said very little to someone who needed to know whether an arrival at eleven at night could be coordinated. We adapt to every guest might sound even better and yet reveal nothing about what adaptation was possible, under what conditions or how far in advance it needed to be requested.
A considerably less spectacular formulation—“If you expect to arrive outside the usual hours, let us know in advance so that we can coordinate your arrival”—could be far more useful because it did not promise unlimited flexibility and brought the signal closer to the fact someone needed to understand.
The smaller the distance between what the hotel knew and what it claimed, the less interpretation someone trying to understand it needed to add.
That precision also mattered when the choice was mediated by artificial intelligence. A system could connect scattered information, compare sources and infer that certain characteristics probably made a hotel suitable for a particular traveller. But the more that conclusion depended on general expressions, missing conditions or claims whose current validity had to be assumed, the more fragile the recommendation became.
Artificial intelligence could enormously expand the ability to connect signals; it could not turn an imprecise signal into precise evidence without adding interpretation.
The responsibility still belonged to the property. If a characteristic was genuinely important to understanding it, the hotel needed to ask whether it had left enough evidence for that characteristic to be recognised without requiring too many assumptions. Not to convince a machine or to write for an algorithm, but because the same precision that helped a person choose for better reasons also helped any system trying to guide them.
That did not mean that every piece of knowledge deserved to become a signal. Some things needed to remain inside because they served the operation, others needed to circulate among people because they helped them make decisions, and some could directly affect what someone needed to know before choosing. The difficulty lay in recognising where that boundary was.
A technical characteristic of a room known to maintenance probably did not matter to any future guest. But if that characteristic consistently affected rest, accessibility or some relevant aspect of the experience, part of that knowledge might need to reach reservations, change a description or influence how the rooms were presented. Something similar applied to a practice learned by reception: it could belong exclusively to the internal operation until fulfilling a public promise under particular conditions depended on it. At that point, those conditions ceased to be merely operational.
Making knowledge visible, then, did not mean publishing everything the hotel knew, but recognising which part of what it knew could change a choice.
That idea also allowed me to look differently at guest reviews. For several conversations, we had used them to discover things the hotel did not say about itself. Now they could also show us where the official information still left a gap. If different guests repeatedly mentioned a capability the operation itself recognised and knew how to sustain, and that capability could matter to future decisions, a different question appeared:
Why should someone discover it only after staying there?
That did not mean appropriating every compliment and turning it into a commercial promise. An extraordinary experience could remain exceptional. But guests’ memories could reveal more than what had happened during a stay: they could point towards real capabilities that the property had not yet managed to make sufficiently understandable before the choice.
I returned then to the person looking for a quiet hotel for a trip with their mother. Perhaps the property we had been observing really was an excellent option. It might have particularly suitable rooms, experience coordinating late arrivals and the ability to accommodate certain dietary restrictions. The team might know all of this perfectly well. If none of it left sufficiently clear signals, however, the recommendation would depend on someone managing to reconstruct it from scattered clues.
An artificial intelligence system could find several reviews, interpret photographs, connect policies and detect consistencies that a person might not have had time to look for. That capability was new and there was no sense in ignoring it, but nor was it sensible to turn it into an obligation to guess. The more important a characteristic was in explaining why a particular guest might find the hotel a good choice, the less reasonable it seemed to leave its understanding entirely to inference.
Nor could the hotel control every conclusion others might draw about it. Experiences would continue to produce different reviews, the media would choose their own angles, platforms would organise information according to their own criteria, and artificial intelligence systems would construct answers using sources the property could not fully direct. What it could do was reduce the distance between what it genuinely knew about itself and the signals it made available, providing sufficiently clear, current and verifiable references around what it considered important.
For years, I had thought of communication primarily as a way of telling people what the hotel offered. Now I was beginning to see it also as a decision about which part of the operation’s knowledge needed to cross the door to help someone who had not yet arrived.
That person might encounter the trace directly—in a policy, a description or a photograph—recognise it in the experiences of other guests, or receive it after a system had interpreted it for them. The paths were different, but the underlying problem remained the same.
The idea needed one qualification, however. Not everything the hotel failed to state explicitly was necessarily lost. Guests could describe it, a photograph could reveal it, a policy demonstrate it indirectly and different experiences allow it to be reconstructed. That was precisely why we had spent so many conversations learning to read signals. The point was not to make the hotel say everything, but to ensure that what truly mattered could be discovered without requiring an unnecessary amount of interpretation.
That was why the word trace began to seem more appropriate to me than message. A message assumes that someone decided what to say. A trace can also appear in a policy, a photograph, an operational decision, a precise description or the recurrence of independent experiences. Some would be produced deliberately and others would emerge as a natural consequence of what the property did. What mattered was that they could reasonably lead towards the reality we were trying to understand.
Perhaps artificial intelligence did not change the hotel’s fundamental responsibility so much as make it more visible. In the past, it might have been enough to appear in front of someone searching for a category, a location or a price. Now a question could simultaneously contain a person, a circumstance, several needs and an expectation. The more precise that question became, the clearer the difference between having information and being understandable.
The hotel could know a great deal about itself and have managed to distribute that knowledge more effectively throughout the organisation. But if what could change a choice did not leave sufficiently clear traces, someone on the outside still had to complete too much of the story for themselves.
And then a consequence appeared that carried the problem a little further. Leaving a good trace once did not guarantee that it would remain true. Rooms changed, and so did policies; a service could disappear, a new capability emerge, an operational restriction change, and what guests remembered could begin to tell a different story. Even the team’s own knowledge could evolve while the public description remained untouched.
The problem would no longer be only deciding which part of what the hotel knew needed to become visible, but ensuring that those traces continued to describe the hotel that actually existed.
Management Perspective
Not everything a hotel knows needs to become public. Some knowledge serves the operation, some needs to circulate among those making decisions, and some can directly affect a future guest’s choice. The latter needs to leave signals that are sufficiently clear, current and close to the reality they describe.
Making that knowledge visible does not mean explaining internal procedures or multiplying messages, but ensuring that what a person—or a system trying to guide them—can understand from the outside corresponds sufficiently with what the property can actually offer.
The Case Study
In the hotel we have been observing, the operation could know with considerable precision which rooms were better suited to particular preferences, under what conditions an arrival outside the usual hours could be coordinated or which adaptations required advance notice. The fact that this knowledge belonged to the hotel did not guarantee that it could be discovered before booking.
The challenge now was to recognise which of those conditions could affect a choice and what traces could make them understandable without turning communication into an operating manual or expanding a specific capability until it became a promise the operation could not sustain.
A Question Before Moving On
If someone needed to understand what your hotel knows how to do particularly well, would they find enough traces to discover it without having to guess?