This is the third and final installment of the series on ChatGPT (the first two are available here and here) and artificial intelligence software now being deployed in the travel retailing space. The pace of change is once again moving at breakneck speed. I have read dozens of articles about actual or potential uses in travel retailing. Every day many new ones are published.
The advent of ChatGPT and its progeny have sparked a reaction comparable to what occurred when airline deregulation was enacted again when commissions were capped and ended (for a time) and yet again when the internet disrupted old ways of doing business. In each of those “revolutions,” there were many conflicting views of the implications and considerable hysteria. We are seeing all that again with AI. Knowledgeable people proclaim a new and better world and knowledgeable people warn that AI represents an existential threat to humanity.
I will not try to resolve those alternative views here. More modestly, the goal here is to sample information, and some opinions, on how artificial intelligence may enter the travel retail space and change how business is done.
A quick disclaimer: I am not a Luddite (a person opposed to new technology or ways of working). But when everyone runs to one side of the boat to see the whale, the boat can capsize. I therefore generally counsel a somewhat skeptical view of “revolutionary technology that will instantly make everything easier and better.” We are usually disappointed by such claims, and AI may be no exception. It’s generally best to stay calm, stay grounded and be careful. So far, ChatGPT itself is not reliable in identifying text produced by AI, including ChatGPT itself!
Just last week an article in this very publication described some uses of AI by travel advisors: 1. Write an advisor biography; 2. Write a Facebook post; 3. Write a blog post.
The suggested key to doing this effectively relies on what are called “prompts.” Prompts frame not only what is to be done but indicate the type of “person” being assigned the task. The example used involved the creation of a legal document related to travel insurance. Color me biased, but I am alarmed that an advisor or agency would rely on such a document without running it by a living attorney with some expertise in travel law. The document created was pretty good, but there are more issues involved in this than the AI covered. The draft might save some legal time, but a “live” review is definitely in order.
By the way, when I asked ChatGPT (as Bing) to do the identical task without the lawyer-persona prompt, the answer was shorter and less comprehensive, more like a disclaimer than an actual legal document. Adding the prompt did not, however, result in a materially more useful response.
The other examples in that article suggest to me that in many cases the uses of AI require more inputs and revisions than would be required by simply assigning the tasks to a knowledgeable human writer.
Conclusion: AI “may” be useful in drafting documents, but it is the beginning, not the end, of the process of creating useful text. Exercise caution.
Another article, reported on using ChatGPT by Hospitable, a software service that helps short-term rental hosts manage properties, to draft responses to guest messages. While praised as “magic,” more than half the AI-generated responses required “small adjustments.”
The Hotels Network, a hospitality technology company, has an AI tool that creates “personalized website messages” in any language. Quadlabs Technologies, an India-based travel technology company, is developing an itinerary builder using ChatGPT that “offers personalized recommendations and allows for booking flights, hotels, and activities in a few clicks…. effortlessly.”
GuideGeek, from Matador Network, claims to be a “global travel assistant” providing “answers to queries about hotels and Airbnbs, custom itineraries, local experiences and tours, even local slang.” It asserts that “the days of spending hours searching Google to plan your trip are over.”
Breezeway, a “property care and services platform, is offering short-term rental property managers a tool to “create suggested replies to guests within the company’s messaging product.” The report was unclear as to whether “suggested” means that managers still must review every message before sending.
Turneo, a startup, plans to “simplify how hotels and other travel brands offer and manage in-destination experiences” by using an AI chatbot “to act as a virtual concierge, offering hotel guests bookable recommendations of local experiences.”
Akin, a hotel operating system, says it has created a ‘personalized, multilingual AI concierge” that “responds to a range of customer-facing interactions, including emails and reviews, with the correct tone and context” that can be enhanced “by reviewing historical conversation logs and providing feedback.” Akin says it is working on software that “processes dynamic rates and availability, enabling guests to make bookings directly with an autonomous agent with only text-based prompts.”
Given that the effective use of AI to generate reliable responses depends to a large degree on the type of prompts input in the task description, I wonder how effective the “average consumer” will be in directing AI bots to provide useful travel information. Time will tell.
The flurry of excited development of AI-based tools for travel has raised privacy concerns, among others. Major technology companies like Microsoft, Adobe, IBM, and Oracle are proposing legislation to address the trust concerns of the new AI capabilities. Trevor Butterworth, co-founder and vice president for communications and governance for Indicio, a developer of “open-source decentralized identity,” is quoted saying, “With the scale of power that’s now available to us, the scale of catastrophe is also raised.”
Meanwhile, as I predicted in an earlier TMR article, development continues uninterrupted. Expedia, Kayak, and Trip.com are among a growing list of travel giants incorporating ChatGPT or its equivalent into online systems.
One proposal is to use “decentralized identity technology,” yet another technological complication with (in my view) associated uncertainty and expense. The new “solution” to search and booking thus begets more complexity and potential risk and cost. There is no free lunch, no matter what the hype claims.
Another expert on “self-sovereign identity [?],” says,
Yesterday it was fake reviews on Tripadvisor, tomorrow it will be you’ve taken my money for a package trip, but I get off the plane and there is no tour guide, there is no bus to collect you, there is no hotel. All the reviews are there but along with the website, the photos, the phone call you had with the agent – none of it is real. AI is surfacing incredible opportunity, but we have a fake everything problem.
Other suggested uses for ChatGPT, or modified versions of it, are:
1. Creating model itineraries for advisor review,
2. Drafting brochure language,
4. Travel information posts/articles on topics like packing lists and how to exchange money.
But even the creators of travel-focused versions of ChatGPT and advisors who have tested them warn that, “It’s important to keep in mind that Oliver [a creation of Travel Research Online] cannot write for you; he [!] is designed to write with you …. It’s important to fact-check everything Oliver says and use a plagiarism checker like Grammarly to make sure you are not copying someone else’s story.”
Doing (Or Not Doing) What Advisors Can Do
One suggestion that caught my eye was to use the AI tool to provide directions from a gate to baggage claim in the airport.” That is a potentially helpful usage (think Frankfurt Airport), but the directions had better be right; errors could result in a client missing a flight.
A recent article by the CEO of Travelpayouts, an affiliate marketing network, reports that AI is now in use in “booking our own flights, checking our own bags, coordinating with hotel staff in languages we don't speak and taking virtual tours of cities we've never been to.” The article notes that “AI is replacing human support teams left and right. And travelers are quickly adapting to these new realities: According to a survey by Ubisend, 48% of consumers don't care whether they get their information from bots or call centers.” Further:
One of the best examples of successful chatbot tech in the travel industry is Juliet by WestJet. Even before Covid, this virtual assistant was able to handle 74% of client support requests. When the pandemic started, Juliet saw a 45-fold increase in requests, but the system easily kept up with demand. That means the technology has essentially been able to do the job of entire client support teams.
Forgive me if my enthusiasm for these developments remains muted. Inevitably, AI enthusiasts are proclaiming that chatbots are replacing travel advisors as the “go-to” sources of travel expertise. One report asserts that:
Sixty-two percent [of study respondents] admitted that they prefer to use an online chatbot service because it quickly can help them instead of waiting for an actual customer service representative to take their call.
ChatGPT can already handle requests like "3-day itinerary in Las Vegas for my 70-year-old parents" with flying colors. Just add a calendar and payment terminal to that functionality and you've got a fully-fledged e-travel agent that can identify the best destination pairs and the most profitable routes and dates by combining over 30 factors in seconds.
TMR readers who were around during airline deregulation, commission caps, and the arrival of the Internet will be familiar with these claims. The author goes on to say that chatbots can “find you the perfect flight” by analyzing consumer data and “personalizing” recommendations based on users’ search histories. Perfect.
Until something goes wrong.
Another recent article reviewed Travel chatbots: Top examples that travelers and businesses love. The first one, WestJet’s Juliet, is just a repeat of another published article. A chatbot called Rose at the Cosmopolitan of Las Vegas claims to “help guests make restaurant recommendations and book tables, spa treatments, event tickets, self-guided tours, as well as answer questions about hotel services.” Its main claimed benefit is that “[Rose’s] playful personality has wooed hotel guests who booked directly into spending 37 percent more than guests who do not engage with her.”
The Edwardian Hotels London’s Edward bot is more interesting. The report says Edward:
teaches guests about hotel amenities, gives directions and tips, and even helps guests submit complaints …. the entire guest experience becomes self-service, including selecting rooms, checking in, and paying.
… Edward can understand with high accuracy guests’ needs and assist with more than 1,200 topics, such as packages included, departure time, upgrades, adding amenities, ordering room service, and more. By integrating with back-end systems, the bot syncs incoming requests with the mobile number to identify the individual guest. Within seconds, Edward accesses the customer profile to provide a hyper-personalized experience. If the bot can’t help, it will route the guest to the specific person at the hotel who can. If a request isn’t taken care of efficiently, Edward will route a request all the way to the general manager.
…. One of the most popular questions Edward gets is, ‘Is breakfast included in my booking?’ Edward will check, and if it’s not included, he will respond with the cost, offer a promo, and add it to the reservation for the guest. Edward can do this much faster than a human needing to look up a reservation.
Edward is responsible for increasing room service sales by 10–50 percent. The bot has helped more than 30,000 guests from 99 countries speaking 59 different languages, and so far in 2019, has managed 69 percent of all guest requests—saving approximately 95 working days.
In my view, aside from the technical wizardry, the emergence of these chatbots means that travel advisors have yet another responsibility – to prepare their clients to deal with these “devices” on-site. A client who is accustomed to or simply prefers to deal with humans may be put off by having a chatbot “intervene” in her service requests. Over time people will become accustomed to dealing with more sophisticated chatbots than the mindless and essentially useless ones we have encountered in the past. In the meantime, the wise advisor will be sure she knows about the presence and capabilities of these new “tools,” so clients are prepared to deal with them.
The other issue, just beginning to get serious attention, is to what extent personal data is going to be secured when an AI-powered chatbot induces a traveler to give up personal data in exchange for service. Aside from the stringent requirements in California’s law, this is a gaping hole in the AI “revolution” that may lead to serious complications in the near future.
Another factor here is the airlines’ objective of taking control of their content. The chosen vehicle is the New Distribution Capability (NDC). NDC has been in development since around 2014 and was supposed to arrive in full very shortly. It’s now 2023 and adoption remains limited. Many articles tying NDC to developments in artificial intelligence continue to present NDC as a kind of miracle solution to a multitude of distribution problems that airlines believe they face.
Finally, ChatGPT has evolved into a paid subscription ($20 a month) for a version that can access the internet in real-time and can accept 70 plug-ins such as Kayak and Open Table. This is just the beginning. Here are some links to recent stories about ChatGPT and related technologies that should be explored by travel advisors. New articles are being published every day and this is a lot of information to take in. Nevertheless, it is imperative that travel advisors understand what is coming and how it can be used, along with the risks involved.
- Magpie Launches New Tool For Tour Operators Built on ChatGPT
- Why Trip.Com Thought It Needed a Generative AI Chatbot
- Ideas from the Yale School of Management -- Management in Practice -- Putting AI on Every Team
- You might start seeing AI chatbots everywhere, thanks to 'the GPT effect'
- A Conversation With Bing’s Chatbot Left Me Deeply Unsettled
- Artificial intelligence like ChatGPT is on the brink of an ‘iPhone moment’ thanks to ‘warp-speed’ development, Bank of America says
- AI in Travel
- After Testing ChatGPT, CNBC Says Travel Advisors’ Jobs Are ‘Secure’ for Now
AI is not coming – it is here. The prospects for meaningful regulation of AI technologies are poor to nonexistent. Advisors therefore should prepare by investigating possible uses.
My final, and repeated, caution is: be careful. Don’t expect AI to have the skills and understanding of a knowledgeable travel expert. It doesn’t and for the foreseeable future, it will not. It’s a tool – an impressive and powerful tool. But, like other tools, in the right hands, it can produce great results.
In the wrong hands, it can be disastrous. It took little time for scammers to figure out robocalling, phone number spoofing, and many other “tricks” to fool people into giving up personal information and/or money in exchange for … nothing. AI is powerful beyond anything seen so far. With power comes danger. One last time – be careful.
Final Final Note: the above article was not written by AI. The named author personally wrote it. He promises. You can trust him. He promises.