ChatGPT: your new travel agency?

Unless you’ve spent the last few weeks on the moon, you’ve probably heard of ChatGPT, the artificial intelligence that’s been making headlines. We could not help but wonder what impact it might have on travel and, in this case, on assisting travellers in organising their trips.

ChatGPT

Without going into technical details, it is a conversational robot based on artificial intelligence. You can interact with it, ask it questions and it answers you based on a knowledge base that it has assimilated.

Its predictive intelligence engine helps it understand what the human is asking and then it searches its knowledge base for the most logical answer and responds in natural language, i.e. by making sentences and talking to you as if it were a human.

ChatGPT vs. Search engine

At this point some might say that a search engine does exactly the same thing and the answer is no! And I will give you some simple examples among others to understand part of the difference.

A search engine gives you links to answers or elements of answers so that you can formulate your answer. ChatGPT gives you the answer (or tries to).

If I want to go on holiday to a country not too far from Europe, where I have both cultural sites and places to relax, I will need to make multiple, well-formulated Google queries to find links to sites where I can find answers, and then I will be free to process this information to choose a destination.

If I ask ChatGPT it will give me a list of countries and tell me why.

With Google I should know pretty much what I’m looking for. ChatGPT can suggest things to me that I wasn’t thinking of.

Google gives me sources in relation to a query, ChatGPT gives me answers in relation to a question trying to understand my intention.

Google does not process the information. ChatGPT does. If one site says that Henry IV has a horse and another that Henry IV’s animal is white, Google will give me a link to each site and I will have to link the two pieces of information. ChatGPT will tell me that Henry IV’s horse is white.

ChatGPT travel agent

The use case I spontaneously thought of was the online travel agent and travel planning. And this is good news, with the inexorable closure of physical travel agencies it is very difficult to get advice on where to go if you don’t have a clear idea of what you are looking for.

I wanted to find a destination that was more cultural but with the possibility of a bit of seaside relaxation, not too far from Europe.

So I asked ChatGPT. The following screenshots are in French, but you’ll have no trouble understanding what it answers to get an overall idea.

He suggests Marrakech, Istanbul, Dubrovnik, Tangier and Lisbon. Not bad but I wanted to go further east.

Ah that suits me better. In particular Amman, which I would not have thought of spontaneously. I therefore ask him for more information on the activities in Jordan.

I am a little disappointed by the quality of the answer. Petra is a world famous site but I think there is much more to do. So I’m surprised by this and say to ChatGPT “is that all?”.

Ah that’s much better. I just have to organise my trip but maybe ChatGPT can help me…

Last concern: what is the weather like in Jordan in March?

A basic example, but one that shows the potential of this technology.

How can ChatGPT enter the world of travel agencies?

Today it is more about experimentation, but tomorrow, in a real business context, how can this technology be offered to the customer?

First of all integrated into a search engine (Microsoft’s Bing is rumoured in the near future). And of course one imagines that the responses could include suggestions for hotels and flights with sponsored links.

Afterwards, it is obvious that the OTAs (online travel agencies) will find it advantageous to include this technology with a real consulting offer that they do not have today. This is a way for them to generate a lot of additional sales and to avoid the customer splitting up his bookings on different sites.

The big losers could be the direct sales channels of hoteliers and airlines with customers succumbing to the ease of the “one stop shop”. Once and for all and for many reasons we can only recommend that you always book directly with hoteliers but there is a real risk that such a facility will divert even more customers from direct channels, which is both detrimental to the hotelier who loses a large part of his margin and to customer who does not pay the best price contrary to popular belief. This is an opportunity for these players to improve the attractiveness of their loyalty programme, which is a real source of direct bookings, and, why not, to start this type of activity. After all, the big chains sometimes have partnerships with airlines and partners for local experiences.

There are four possible entry points:

– An OTA, a hotel group or an airline integrates it into its online offer.

– A business creates a new product and works to connect ChatGPT with distribution and reservation systems and offers the technology to the latter as a white label

– A distribution and booking company like Amadeus integrates the technology into its products.

– A pure player is created.

The limitations of ChatGPT

The main limitation of ChatGPT is the body of knowledge available to it. Today it is limited and sometimes dated: the information it uses today is prior to 2021 and it is not (yet?) connected to the internet.

If I were to ask ChatGPT about the formalities required to enter Jordan I doubt this information would be up to date given the speed at which it changes.

ChatGPT does not have access to inventories so there is no need to ask about prices and availability. Using it in the use case I have presented will require integration with GDSs.

On the other hand, I was quite surprised by his answers to a few questions, proving that his body of knowledge is not so small.

I asked him to advise me on the best airline to go from Paris to Amman.

It and I do not agree on the notion of the “best” airline, but that’s another debate.

Same question for finding a luxury hotel in Amman.

Relevant and argued but rather obvious answers. To try to understand where these answers come from and how reliable they are, I ask him about one hotel in particular: the W Amman.

Je dois admettre que la réponse est argumentée et crédible.

The earlier WayBlazer

This use case of ChatGPT seemed obvious to me and for a good reason: I had already analysed the same thing 7 years ago with Wayblazer, a service based on IBM Watson AI and the IBM case study explains very well how it works and the benefits to be expected. The operating principle is exactly the same.

And the results were there.

But Wayblazer went out of business in 2018. Should we draw any conclusions from this for ChatGPT? We don’t think so.

The technology was in its infancy and expensive at the time, certainly much more expensive than today and more complex to implement. And even though IBM communicated a lot about Watson there was not the current hype, and certainly the market was less mature and demanding of such uses than today.

One would like to think that Wayblazer led the way but arrived too early with a technology that, although disruptive at the time, was still in its infancy.

Bottom line

ChatGPT has the potential to transform the world of travel agencies and travel planning by bringing online and on a large scale an advisory service that the customer could only get in the agency, often reserved for a somewhat premium clientele.

And unlike its predecessor Wayblazer, we think the market is ready for it.

But for this technology to deliver its full value it will need to be connected to the various distribution and booking engines and this will be the challenge for anyone entering the market. And not the least.

Image : ChatGPT by Ascannio via Shutterstock