ResLobby
If you're searching "hotel chatbot", you're probably at this point: the incoming messages are piling up, most are the same questions, and you heard someone has automated this.
You're in the right place. But you're probably searching for the wrong word.
In this article I'll explain what a chatbot is, where it works, where it falls short, and what the alternative is. In the end there's also a scenario where a chatbot is enough for you — and I'll say that too.
What is a hotel chatbot?
A hotel chatbot is software that answers guest messages automatically according to pre-defined rules. It recognizes specific keywords or menu choices and sends the ready-made text corresponding to each. The flow of the conversation is set by a diagram drawn by hand during setup. When a question arrives that isn't in the diagram, the chatbot cannot produce an answer.
That last sentence is the subject of this whole article.
When is a chatbot enough?
Let's be honest first: not every business needs a more advanced system.
A chatbot is enough in these cases:
- If almost all incoming questions consist of five or six fixed questions
- If most of your guests are used to picking from a menu
- If the cost of a wrong answer is low
- If your information rarely changes
A chatbot that tells a small cafe's opening hours is perfect. It has one job to do, and it does it without error.
Accommodation isn't that simple. And here's why: a guest asks questions when buying a room; they don't when buying a coffee.
Where does a chatbot fall short?
1. It understands the word, not the sentence
If the chatbot is waiting for the word "price", it gets stuck on "How much is a double room?"
The guest asking "Can I bring my dog?" didn't use the word "pet". The chatbot misses this.
The question "Can you walk to the sea?" doesn't match the "distance" scenario.
Every time, the guest tries to find the word the chatbot is waiting for. But the guest doesn't know your scenario. They speak naturally.
2. Every new scenario is written by hand each time
You set up a chatbot, defined fifteen scenarios. Two weeks later a question no one thought of comes in: "Do you have a baby cot?"
It's not in the scenarios. The chatbot can't answer.
You open the panel and add a new scenario. Then another. Then another.
A year later you have ninety scenarios and even you no longer know which question goes to which flow.
This work has a low setup cost and a high maintenance cost.
3. It's unclear what it'll do on a question it doesn't know
This is the most critical point.
What does a chatbot do when it can't answer a question? Mostly it either returns to the main menu ("How can I help you?"), stays silent, or triggers the nearest scenario — which means a wrong answer.
The guest is asking about a group of eight. The chatbot triggers the "family room" scenario. The guest gets an irrelevant answer and closes the conversation.
No one measures this, because the loss is silent. The guest doesn't complain, they just leave.
Chatbot or AI customer assistant?
They are two different categories. The difference isn't just how natural the conversation feels.
| Feature | Rule-based chatbot | AI customer assistant |
|---|---|---|
| Understanding the question | Keyword / menu | Meaning of the sentence |
| Source of information | Hand-written scenarios | The business's knowledge base |
| New question | A scenario is added by hand | Already understood |
| Unknown question | Returns to menu or answers wrong | Asks the business owner |
| Maintenance | Constant scenario maintenance | Learns as information is added |
| Risk | Irrelevant answer | Wrong information (unless prevented) |
The last row matters. Both categories carry a risk. A chatbot gives an irrelevant answer. An AI customer assistant, if poorly designed, makes things up.
We can't move on without discussing this.
So doesn't an AI assistant lie?
It can. This is called hallucination.
If it doesn't know whether the hotel has a room with a jacuzzi, it might write "Yes, we have jacuzzi suites available." It sounds natural, fluent, convincing — and it's wrong.
In hospitality the cost of this is concrete: the guest arrives, there is no room, the reputation takes a hit.
This isn't an unavoidable flaw of the technology. It's a design choice.
A well-designed system does this: when a question with no match in the knowledge base comes in, it does not guess. It admits what it doesn't know and forwards the question to the business owner. The owner writes the answer, it reaches the guest, the system learns.
At ResLobby this mechanism is called Smart Handoff. A notification drops into your Telegram, you answer, and the same question is never asked again.
So the right question isn't "does AI make things up?" The right question is: "What does this system do when it doesn't know?"
That's the only question to ask when evaluating a product. If the answer isn't clear, that product isn't ready.
Can a hotel chatbot take reservations?
Technically it can. It should not.
Here's why: no automatic system can know current availability for certain. Your reservation book, requests coming by phone, rooms canceled but not yet entered into the system — none of these are real-time.
A wrong confirmation leaves you alone with a guest waiting at the door.
The correct behavior is this: when the topic of a reservation comes up, the system directs the guest straight to you. It gives the phone number, connects to WhatsApp. It opens the door, it doesn't make the decision.
ResLobby works this way. Price and availability questions are in the same category — the assistant doesn't guess, it brings you into the loop.
Cost: not setup, but maintenance
Chatbot prices are usually discussed in terms of setup. Wrong metric.
The real cost emerges after the first year:
- Hours spent adding new scenarios
- Finding flows that give wrong answers
- Updating seasonal changes by hand
- Missed guests (the invisible cost)
In the knowledge-base approach this cycle works differently. You add a sentence to the knowledge base, the assistant uses that information. You don't draw scenarios.
We looked at why this burden is heavier at small properties in a separate article.
Three questions to decide
These three questions determine which category you need:
1. How many different questions do your guests ask?
Five or six, a chatbot is enough. Twenty or thirty, scenario maintenance will wear you down.
2. What is the cost of a wrong answer?
Low at a cafe. High at a hotel. A guest who arrives on wrong information means a lost review.
3. What will the system do when it doesn't know?
Clarify the answer to this before buying. "Returns to menu" is one answer. "Asks you" is another. The difference between the two is the product's real value.
If you're searching "hotel chatbot", what you're actually looking for is probably that guest questions get answered correctly. A chatbot is just one of the tools that does this — and for hospitality, often the weakest.
If you'd like to see how ResLobby works at your property, you can reach us through the Get Started button on the homepage.
Related articles:
This article is part of the guide: Hotel Guest Communication: A Complete Guide