GPT in tourism and hospitality.
Legacy article: written for GPT-3, no longer current. Kept for the record.
The whole premise of GPT-3 is to make services more efficient without lowering their quality. That’s particularly good news for tourism and hospitality businesses, which run on highly complex systems. Don’t be shocked when GPT-3 plans your next trip to the Mediterranean.
“Quality is more important than quantity. One home run is much better than two doubles.” Steve Jobs
“Quantity has a quality of its own.” Napoleon Bonaparte
“I can do both well :)” GPT-3
GPT-3 arrived at the perfect time. The industry faces a lot of challenges, not least the struggle to maintain good customer service: sometimes because of high demand, sometimes because of labour shortages, and cruelly, sometimes both. Here GPT-3 helps companies by handling direct and simple questions, analysing data, offering a creative angle when writing content and much more. That frees employees to deal with the complex tasks.
Travellers, guests and tourists like to stay in touch with the company throughout the whole customer journey. In the dreaming, researching and planning phases they want as much information as possible, to reduce uncertainty and the risk of a bad decision, which can cost a lot of money. That’s because travel services can’t be tested in advance: they are produced and consumed in a single act. So a lot of contextual content has to be produced for several touchpoints. After the trip, on the other hand, reviews and user-generated content dominate. Here, proper analysis can bring good insight into customer sentiment, and GPT-3 does a decent job of it.
Before getting too excited: GPT-3 still can’t do everything, and it’s better that it doesn’t. But it has proved to perform well in several cases.
Customer service
GPT-3 can be trained to take a large number of variables into account. These can be (1) external: number of rooms, rate, location, distance to the airport; or (2) travel preferences: preferred type of hotel or favourite season to travel. That makes the response more personalised, more like a human’s. With that knowledge, GPT-3 can answer a lot of common questions and also suggest an action. For example, if a traveller wants to know the cheapest hotel on 15 November, GPT-3 can give the answer and offer to book it. The same applies to refunds or room availability. Amadeus recently partnered with Aimendo to automate communication between agents and travellers.
Content creation
By content I mean blogs, but also emails, seasonal offers and hotel descriptions. What’s fascinating is that GPT-3 can adapt to different tones of writing and levels of creative depth. Maybe you want the tone of your next email to be a little funnier. You can do that. I’ve read people suggesting multiple tones for multiple target groups. I think that’s dangerous: it can confuse the reader, and it’s open to category bias. Can’t a business traveller also be a family traveller? Just because you can doesn’t mean you should.
GPT-3 can also be a lifesaver if you want to translate content into several languages. A hard hit for translators, isn’t it? Mack Grenfell has built a tool, byword.ai, that lets you generate high-quality articles at scale. I’ve used it to write some of the travel articles on my blog, traveltokorca.com. These tools can also optimise texts for particular keywords, including long-tail ones.
Data analysis
First and foremost, GPT-3 can help marketers understand customer sentiment. This is powerful. We can now extract all the reviews from third-party websites (Tripadvisor, Booking.com, Expedia) and run a simple sentiment analysis. What do customers think of us? Is their opinion positive, negative or neutral? We can use that to improve customer service in future. If you want to know exactly how to do it, Michael Taylor (Vexpower) has built a free course on it. Second, GPT-3 can generate countless ad copy variations for you to test.
One thing should be clear, though. We must decide carefully how much power we delegate to GPT-3 and other tools. Should it stick to tactical work, or is it valuable in strategic planning as well? Get that wrong and it’s like outsourcing key strategic capabilities to an external agency.
Update: things are moving fast. ChatGPT, text-davinci-003.
Thanks to
- Michael Taylor
- Colin Lewis
- Roman Egger
- Mack Grenfell