How You Talk to ChatGPT Says More About Your Leadership Than You Think

Yesterday, I caught myself thanking ChatGPT. Nothing extraordinary had happened. I had asked it to review an idea, it had responded reasonably well, and I ended up writing a polite thank you, almost as one would when saying goodbye to someone who has just helped them at Reception. Then I found myself staring at the screen, wondering: why am I thanking a machine? And, while I was at it, why do I sometimes address it informally, sometimes adopt an almost legalistic tone, and on occasion come dangerously close to arguing with it as if it had changed a reservation without telling me?

The question may seem frivolous, but it goes deeper than it appears. How do you speak to ChatGPT? Do you ask for things politely? Do you say good morning? Do you use a formal register when the request is serious and an informal one when you simply want to summarise a document? Do you get angry when it does not understand an instruction? Do you write in capitals? Do you insult it? Do you thank it when the result helps you? I suspect that each of us has gradually developed our own etiquette protocol with Artificial Intelligence, even though nobody has handed us the corresponding guide to good manners yet.

After many years in Hospitality, I have learned to pay close attention to language. In a hotel, the way something is requested is rarely an innocent detail. The same instruction can generate commitment, confusion, fear, indifference or initiative, depending on how it is phrased. Content matters, but so do tone, context, timing and the respect we convey. That is why I am less interested in whether ChatGPT deserves our thanks than in understanding what the way we speak to it reveals about how we think, request, correct and lead.

It is worth clarifying something before anyone begins worrying about the algorithm's feelings. ChatGPT does not leave the conversation feeling crushed because you treated it badly, it does not discuss your bad mood with another model during a coffee break, nor does it request a meeting with Human Resources because you wrote five exclamation marks. Courtesy is not necessary to emotionally protect the tool. However, that does not mean our manners are irrelevant. The language we use also shapes us, and it can end up carrying over into conversations where, on the other side, there is indeed a person.

My reflection, therefore, is not intended to establish whether AI should be addressed formally or informally. Both options can be perfectly valid. What is truly useful for a Hospitality professional is to ask whether this new everyday relationship is improving their ability to give instructions, organise ideas, accept corrections and maintain judgement under pressure. Because perhaps ChatGPT does not need our good manners, but we do need to preserve them. And, in the process, we can use every conversation to review how we are communicating with our teams, our guests and even ourselves.

The machine does not take offence, but your communication habits still matter

I admit that my register with ChatGPT changes depending on the moment. When I am exploring an idea, I address it informally. When I am preparing a complex analysis, my tone becomes more formal. If I am tired, I write shorter instructions. If I have been working on an issue for some time, I tend to assume that the tool knows a context that I have not actually provided. And when the outcome does not fit, my first reaction may be to think that it has misunderstood, although I often discover that the problem was not the answer, but the quality of my request.

This situation feels uncomfortably familiar. In hotels, we also ask for results while assuming that others know all the information we have in our heads. We assign a task in two sentences, omit priorities, fail to explain what decision needs to be made, and then are surprised when the execution does not match our expectations. With ChatGPT, we tend to call it a failure of interpretation. With a team, we sometimes call it a lack of initiative. The difference in judgement is rather convenient for the person who gave the instruction.

AI Hotel Intelligence

AI is only as smart as the context you give it. HotelGEX gives it your whole hotel.

Guests, Revenue, operations, F&B, Groups & Events and Management connected in one intelligent context, so HotelGEX AI can understand what is really happening across your hotel.

Discover HotelGEX AI → Start 30 days free

HotelGEX Intelligence

AI

One AI.
Full hotel context.

Intelligence that connects what your hotel already knows.

Understand

Anticipate

Recommend

AI across the hotel

Customer Intelligence Revenue AI Operations AI F&B Intelligence Groups & Events AI

The way we speak to AI can therefore become a kind of professional mirror. A mirror without emotions, hierarchy or fear of challenging us when we ask it to do so properly. By observing our conversations, we can identify habits that also affect hotel management, team coordination and the guest experience in hotels.

• Using an informal or formal register defines a tone, not a hierarchy. I do not believe there is one correct formula. Informality makes for a direct, natural conversation; a formal register can help us adopt a more analytical tone. What matters is understanding that the tool generally adapts to the language it receives. If we write vaguely, excessively or grandiloquently, we are likely to get something similar in return. If we use a clear, calm and professional tone, we increase the chances of receiving a useful response. In Hospitality terms, we might say that AI does a certain amount of matching with the guest too, although we have not yet taught it to spot who comes down for breakfast before the doors open.
• Saying please does not replace explaining things properly. We can write the most polite prompt in the world and still receive a useless answer if we do not provide context, objectives, constraints and format. Courtesy improves the atmosphere of a human conversation, but it does not fix a flawed instruction. Something similar happens in Hospitality: politely asking someone to prepare a commercial proposal does not clarify the target segment, margin, dates or operational capacity involved. Kindness is necessary, but it cannot be used as an excuse for a lack of precision.
• Saying thank you can be useful even if the machine does not need it. I still do it sometimes. Not because I think the system will remember and offer me a room with a better view in the next conversation, but because thanking is part of the way I close an interaction. Habits of respect are maintained by practising them. When we stop distinguishing between conversations with people and conversations with systems, we risk normalising dry, impatient and purely transactional communication. And Hospitality cannot afford to turn all its relationships into service orders.
• Getting angry with ChatGPT usually reveals a poorly managed expectation. The tool can make mistakes, invent connections, misinterpret an instruction or produce a mediocre response with admirable confidence. It is worth correcting it and verifying what matters. But it is also healthy to review what we expected from it. If we ask it to decide for us, resolve a human conflict, design a strategy without data or automatically understand the particularities of our hotel, perhaps the error began before we pressed send.
• Insulting AI does not harm AI, but it rehearses a way of reacting. Someone may think it does not matter because nobody is receiving the insult. Technically, that may be true. Professionally, I find it more debatable. Self-control is not a garment we put on exclusively for meetings. It is a capability that strengthens or weakens in small moments. If our automatic response to a poor result is aggression, even towards a screen, we are practising a dangerous sequence: frustration, contempt and release. In hotel operations, that sequence will sooner or later find a recipient of flesh and blood.
• Apologising to ChatGPT may reveal something even more interesting. Some people write sorry, I did not explain myself clearly when they rephrase a request. The machine does not need that apology, but the phrase contains a valuable skill: recognising that the problem may lie in our own communication. In an industry full of urgency, crossed interpretations and incomplete messages, that small willingness to review one's own responsibility is worth far more than any prompt manual.

There is also an interesting contradiction. We can be extremely polite to a tool and very inconsiderate towards the team. We can also speak abruptly to AI while maintaining impeccable professional relationships with people. I am not suggesting that we diagnose hotel leadership by examining whether someone writes please in ChatGPT. It would be a rather entertaining methodology, but scientifically dubious and likely to lead to meetings that would be difficult to explain.

What I am suggesting is that we use these conversations as an opportunity for observation. When nobody is watching us, when the interaction has no social consequences and when the other party cannot be offended, a fairly spontaneous version of our communication style emerges. That is where we see whether we organise our thoughts before making a request, whether we vent frustration, whether we know how to correct without drama, and whether we accept that a poor answer may stem from a poor question.

In hotel culture, words never travel alone. They are accompanied by position, urgency, background and emotions. A manager who writes this is wrong, do it again may believe they are being efficient. For the person receiving the message, too many questions remain open: what is wrong, what should be retained, what result is expected, when must it be delivered, and can they ask for help? Replace ChatGPT with a newly hired employee and the situation is no longer amusing.

AI allows us to see, at very low cost, how precision affects outcomes. If we rephrase an instruction by adding context, examples, boundaries and criteria, the answer usually changes. That experience should make us more humble when reviewing a team's performance. Perhaps some people do not need more pressure, but rather a better-designed briefing.

Speaking to ChatGPT as a hotelier: less ceremony, more judgement

There is no need to turn every conversation with ChatGPT into a diplomatic audience. Nor would I recommend beginning every prompt with good morning, asking about its algorithmic family and signing off by wishing it a pleasant day of processing. Courtesy can be natural, but the professional objective remains to obtain useful support without delegating what is ours to think through, decide and take responsibility for.

When I use ChatGPT to work on Hospitality matters, I try to imagine that I am giving instructions to an intelligent, fast person with extensive general knowledge, but who has just joined the hotel and knows almost nothing of what we take for granted here. It does not know our value proposition, the building's constraints, the actual guest profile, the restaurant's capacity, conflicts between departments, commercial commitments or those operational exceptions that never made it into the procedure because everyone thought they were obvious.

This way of understanding the tool changes the conversation. Instead of simply asking it what I should do, I give it a clearly defined problem and ask it to help me examine it. I am not seeking an artificial authority to confirm my intuitions. I am looking for perspective, structure, hypotheses and questions I may not have considered. The difference may seem minor, but it separates superficial use from genuine support for thinking.

I have gradually developed a way of working that can also be applied to meetings, delegation and training processes in the hotel. It is not a magic formula, nor does it aspire to compete in solemnity with the countless methods that appear every week. It is simply a discipline of conversation.

• Start by explaining the situation before asking for a solution. Describe the type of property, the segment, the operational moment, the observed problem and the relevant constraints. Analysing a drop in conversion at a leisure resort is not the same as analysing one at an urban hotel with a high dependence on corporate business. Nor is proposing a breakfast improvement the same when there is spare capacity as when the kitchen is already operating at its limit. Context prevents generic answers and, in real life, also prevents decisions that ignore operations.
• Define the outcome you need. Sometimes we want ideas; at other times, a critical assessment, a table, an action plan, scenario modelling or a risk review. Simply asking to analyse this is the equivalent of walking into Reception and saying sort it out. Someone may well manage it, but we will have made ambiguity part of someone else's work. A good briefing reduces that waste of attention.
• Set out the boundaries that must not be crossed. In hotel marketing, for example, a proposal may seem brilliant and yet be incompatible with the positioning. In hotel revenue management, an action may increase occupancy while damaging margin, direct channel performance or perceived value. In operations, an appealing idea may shift an impossible workload onto the team. Asking for economic, commercial, cultural and operational boundaries to be respected improves the quality of the analysis.
• Ask it to challenge your assumptions. This is one of the most valuable instructions. If we only use ChatGPT to confirm what we already think, we will have symbolically hired the most accommodating adviser on the market. It is worth asking for counterarguments, risks, side effects and alternative explanations. Hotel strategic planning improves when we subject our convictions to a degree of friction before turning them into decisions that other departments will have to bear.
• Correct with precision, not anger. If the result is poor, merely pointing out that it is poor contributes very little. Explain which part does not work, which assumption is incorrect, what depth is missing or what tone does not fit. This practice has a direct application in leadership in the hotel industry. Useful feedback describes the gap between the result received and the expected result; anger only describes the emotional state of the person evaluating it.
• Verify before using. A well-written answer may contain errors, oversimplifications or unsuitable recommendations. A convincing tone is not a guarantee of truth. In financial, legal, employment, commercial or reputational matters, professional review remains essential. AI can help us think, but responsibility cannot be outsourced. If a hotel strategy goes wrong, explaining to ownership that the algorithm seemed very confident is unlikely to improve the meeting.
• Say thank you if it is part of the way you relate to others. It will not magically improve the next result or open an account for digital courtesy points. Yet there is no serious reason to abandon a kind habit simply because the other party is a machine. I prefer to preserve some human rituals, especially when the environment pushes us to communicate ever faster, with less context and less consideration.

This approach has implications that go beyond learning to write good prompts. ChatGPT is introducing many professionals to a new kind of interaction in which the quality of the outcome depends, in part, on the quality of the assignment. And that should interest us enormously in Hospitality, because our hotels operate through thousands of daily assignments: preparing a room, responding to a review, amending a forecast, reorganising a shift, resolving an incident, explaining a rate, designing a campaign or rebuilding a guest's trust.

If a team learns to formulate better questions for AI, it may ultimately learn to ask better questions of one another. What are we trying to achieve? What information is missing? What constraints exist? Who will be affected? How will we know whether it has worked? What risk are we overlooking? These questions strengthen judgement and prevent innovation in Hospitality from becoming a succession of quick answers to problems we have not yet understood.

I have also found that the tone used with ChatGPT influences the tone it returns. This is particularly relevant when the tool is used to draft guest communications, review responses, commercial messages or hotel marketing content. If the prompt is loaded with irritation, defensiveness or a need to prove that the hotel is right, the final text may retain some of that tension, even when it is wrapped in apparently polite phrases.

A response to a complaint can include all the appropriate formulas and still convey hostility. The same happens with certain internal messages: they begin with dear team and, three lines later, have already arranged a public execution with half the company copied in. Formal politeness does not compensate for punitive intent. That is why, before asking ChatGPT to draft a response, it is worth asking ourselves what we are trying to achieve: protect the hotel's ego, close the matter, win back the guest, explain a boundary or learn from an incident.

In the hotel guest experience, language has an economic impact. Confusing communication can lead to calls, repeated contacts, cancellations, incorrect expectations and additional work. A defensive response can turn a recoverable incident into a negative review. An aggressive commercial message can destroy trust. Speaking well to AI is not merely ornamental when that AI helps us produce texts that guests, owners, suppliers or employees will later read.

There is also a straightforward application for team training. We can ask several people to frame a query about the same hotel issue and compare the responses they receive. Not to reward the author of the most sophisticated prompt, but to observe how each person structures the matter. One will include guest information, another will focus on costs, another will identify operational constraints, and perhaps someone will forget to explain the objective. The comparison reveals different ways of understanding the hotel and opens a far more valuable conversation than a theoretical class on Artificial Intelligence.

I would carry out this exercise using an everyday case: a family requests to keep two rooms after check-out time on a high-occupancy day. I would ask the team to use ChatGPT to consider alternatives. I would then review what information each person provided: the time of new arrivals, inventory status, Housekeeping availability, guest value, the cost of blocking rooms, the possibility of offering just one room, access to facilities, luggage storage or reputational impact. The quality of the prompt would reveal the breadth of the hotel thinking of the person who wrote it.

The interesting point would not be discovering who obtained the most elegant answer, but who understood the system best. In Hospitality, decisions rarely belong to a single department. A commercial concession affects Reception, Housekeeping, Revenue, Maintenance, Food and Beverage and, on occasion, other guests. ChatGPT can help us visualise those connections, provided we are capable of including them in the conversation.

There is another, less comfortable lesson. Some people speak to AI with far more patience than they show their colleagues. They rephrase an instruction to ChatGPT three times, add context, provide examples and try different approaches. Yet when an employee does not understand something the first time, they conclude that the person is not paying attention. The machine receives pedagogy; the person receives impatience. Perhaps because we expect little from the former and too much from the latter.

The opposite also happens. Some people spend twenty minutes metaphorically insulting the system over a poor response when they could have rephrased the query in two. That behaviour recalls certain operational conflicts in which we invest more energy in finding someone to blame than in accurately describing the problem. Anger may offer immediate satisfaction, but its return in Hospitality is rather poor.

That is why I believe that the way we speak to ChatGPT can become a small exercise in leadership. We can practise clarity without arrogance, high standards without contempt, correction without humiliation and critical thinking without the need to demonstrate superiority. The tool will not necessarily learn a moral lesson from that behaviour. We might.

The next time you open ChatGPT, I suggest a simple observation: before sending the message, read it as if it were addressed to a competent person who has just arrived at your hotel. Check whether they would understand the context, the objective, the boundaries and the expected outcome. Then ask yourself whether the tone invites them to think or merely demands obedience. That small exercise can improve the prompt, but it can also teach you something about how you are communicating priorities in the operation.

You do not need to address AI formally or ask permission to consult it. Nor do you need to remove the please or thank you to prove that you understand how it works. Speak in the way that feels natural to you, but do so with clarity, judgement and self-control. Courtesy towards a machine is optional; the discipline of communicating well should be mandatory for anyone who works with people.

And if one day you find yourself arguing with ChatGPT because it has failed to understand your instruction for the fifth time, pause before declaring war on Artificial Intelligence. Review the prompt, add the missing context and try again. If it still responds poorly, you can close the laptop gracefully. In Hospitality, we already have enough difficult conversations without adding a personal dispute with an algorithm to the list.