Quick viewing(Text Mode)

Travel Chatbots Are Hot, but Will They Replace Or Augment Human Conversation? by Norm Rose September 2016

Travel Chatbots Are Hot, but Will They Replace Or Augment Human Conversation? by Norm Rose September 2016

ANALYSIS

Travel Chatbots are Hot, But Will They Replace or Augment Human Conversation? By Norm Rose September 2016

In the last several years, conversational bots – computer programs that enable text conversations – have been among the hottest technology topics. Startups and traditional travel players have launched a variety of chatbots focusing on customer services, trip planning and booking. This analysis reviews the current state of chatbots and how they are being implemented by travel companies, and discusses the future impact of this technology on the travel search­shop­buy experience.

This content is published by Phocuswright Inc., a wholly owned subsidiary of Northstar Travel Media, LLC.The information herein is derived from a variety of sources. While every effort has been made to verify the information, the publisher assumes neither responsibility for inconsistencies or inaccuracies in the data nor liability for any damages of any type arising from errors or omissions. All Phocuswright publications are protected by copyright. It is illegal under U.S. federal law (17USC101 et seq.) to copy, fax or electronically distribute copyrighted material beyond the parameters of the License or outside of your organization without explicit permission.

© 2016 Phocuswright, Inc All Rights Reserved. Travel Chatbots September 2016

About

Phocuswright is the travel industry research authority on how travelers, suppliers and intermediaries connect. Independent, rigorous and unbiased, Phocuswright fosters smart strategic planning, tactical decision­making and organizational effectiveness.

Phocuswright delivers qualitative and quantitative research on the evolving dynamics that influence travel, tourism and hospitality distribution. Our marketplace intelligence is the industry standard for segmentation, sizing, forecasting, trends, analysis and consumer travel planning behavior. Every day around the world, senior executives,marketers, strategists and research professionals from all segments of the industry value chain use Phocuswright research for competitive advantage.

To complement its primary research in North and Latin America, Europe and Asia, Phocuswright produces several high­profile conferences in the United States and Europe, and partners with conferences in China and Singapore. Industry leaders and company analysts bring this intelligence to life by debating issues, sharing ideas and defining the ever­evolving reality of travel commerce.

The company is headquartered in the United States with Asia Pacific operations based in India and local analysts on five continents.

Phocuswright is a wholly owned subsidiary of Northstar Travel Media, LLC.

© 2016 Phocuswright Inc. All Rights Reserved. 2 Travel Chatbots September 2016

Table of Contents

Abstract 5 Introduction 5 What are Next­Generation Bots? 5 Bots on Messaging Platforms 6 Bots, Humans or a Hybrid Model 6 Are Chatbots an Interim Step to Voice­Enabled, Human­to­Machine Conversations? 7 Chatbots and the Impact on Search­Shop­Buy 7 Examples of Travel Bots and Human­Driven Texting Solutions 8 New Leisure Travel Services 8 OTA 9 Corporate/Business Travel Services 10 Hospitality 10 Metasearch 11 Airline 11 Ground Transportation 12 Summary 12 Endnotes 13

© 2016 Phocuswright Inc. All Rights Reserved. 3 Travel Chatbots September 2016

Figures and Charts

Fig. 1: Growth of Messaging Platforms 6

© 2016 Phocuswright Inc. All Rights Reserved. 4 Travel Chatbots September 2016

Abstract

The phrase "I've got an app for that" is quickly being replaced by "I have a bot for that." Since late 2015, conversational bots, computer programs that enable a conversation via text, have become one of the hottest technology topics. Startups and traditional travel players have launched a variety of chatbots focusing on customer services, trip planning and booking. The main driver of the chatbot trend is simplicity. Using an existing messaging service, a chatbot can help automate tasks via text without the need to download an app. Some believe that chatbots represent the beginning of a post­app world. This analysis reviews the current state of chatbots and how they are being implemented by travel companies, and discusses the future impact of this technology on the travel Search­Shop­Buy experience.

Introduction

2016 is already being called the year of the bot. MobileBeat 2016, a conference sponsored by Venture Beat, was devoted to the subject of Bots and Artificial Intelligence (AI). The conference lasted two days and had over 30 speakers, with strong representation from the travel industry. A number of questions were raised at the conference. Can bots truly hold a conversation? Are bots simply a stepping stone toward AI­powered human­to­machine voice interfaces, or will the bot phenomenon continue to have a distinct role within text­based conversations? How are travel companies using bots today, and what is the future of bots in our industry? What is the best mix of bots and human­based messaging?

What are Next­Generation Bots?

Conversational bots, which are also called chatbots, are not new. The chatbot concept was introduced in 1966 by MIT computer scientist Joseph Weizenbaum. He created a program for mimicking human conversation called ELIZA, which worked by parsing the words that users entered into a computer and then matching them to a list of possible scripted responses.1 The limits of ELIZA were based on the knowledge contained in the scripts. Though technology has advanced significantly since 1966, the depth of knowledge to answer all possible questions continues to be a challenge for even today's most sophisticated chatbots.

The 1990s saw a first wave of chatbots. Early bots were based on simple logic tree applications (SLT), which gather information and redirect the user to pre­determined answers. When interfacing with an SLT chatbot, a seamless experience can happen if the query matches a pre­determined answer, but questions that stray from the predicted responses often hit a dead­end.2 As a result, the 1990s SLT chatbots failed to produce true conversational capabilities.

Evaluating the conversational capability of a chatbot relates directly to the current state of AI. The Turing test has long been a measure of the effectiveness of AI and can generally be tested in a chatbot context. The test was developed by English mathematician Alan M. Turing to determine whether a computer can appear to think like a human brain. The idea is simple: if the researcher cannot detect that an electronic conversation with a computer masquerading as a human is not actually a human, the software passes the Turing test.3 Today the effectiveness of chatbots is still often measured against the Turing test.

Today, bots reflect the convergence of three types of AI disciplines: Natural Language Processing (NLP), Machine Learning and Cognitive Computing. Natural Language Processing is the ability of a computer program to understand human speech or text. More recently, NLP has been replaced with the term Natural Language Understanding (NLU), which more specifically implies that the machine understands the request. NLU generally requires an ontology (e.g. a set of common terminology with a specific discipline) that defines terms within a specific sector. For example, texting a chatbot to book a flight might yield a response that describes books about flight vs. actual options for air travel. Defining the ontology of the travel industry is critical to move NLP to NLU. NLU applies to both voice and text communication and is now used extensively in conversational bots.

Machine Learning, as the name implies, is the ability for a computer program to learn patterns based on usage. For example, an airline booking platform that utilizes Machine Learning may suggest specific flight times or carriers based on past behavior

© 2016 Phocuswright Inc. All Rights Reserved. 5 Travel Chatbots September 2016

of the user.

The term Cognitive Computing often encompasses both NLU and Machine Learning, but adds Big Data analysis to the application analyzing large amounts of structured and unstructured data to deliver intelligent recommendations. IBM's Watson platform is the most prominent example of cognitive computing. 4

The current wave of chatbots that utilize NLU, Machine Learning and in some cases Cognitive Computing deliver responses that make sense within the context of the conversation. Despite the use of these AI techniques, there is still significant debate on whether conversational bots can handle the complexity inherent in the Search­Shop­Buy process of travel.

Bots on Messaging Platforms

One of the key differences between the 1990s bot revolution and today's chatbots is the mass adoption of messaging platforms. For example, Facebook Messenger opened up their Messenger platform to bots earlier this year, and a mere 80 days after the announcement, the network was flooded with more than 11,000 chatbots.5 The growth of other messaging platforms such as WhatsApp, WeChat, Instagram and Snapchat, in addition to the traditional SMS (e.g. mobile text messaging), has changed the way chatbots are being deployed. Rather than requiring consumers to enter a private chat session via an app, the majority of today's chatbots are being deployed on top of these robust messaging platforms enabling conversations without requiring the user to leave their preferred messaging environment.

Figure 1

In addition, enterprise messaging platforms such as Slack also provide a new platform for chatbots. Slack's growth has been phenomenal, recently reaching 2.7 million daily users.6 A number of travel chatbots have deployed on Slack's group messaging environment.

Bots, Humans or a Hybrid Model

For decades, science fiction novels, television shows and movies have prepared us to interact with intelligent machines. The use of AI, the ability for machines to mimic human behavior, is commonplace in and outside the travel industry. In fact, when AI becomes mainstream, we often take it for granted and don't think of it as AI. It ranges from the obvious — personal assistants such as Siri, Google Now and Cortana — to the use of AI in video game characters that learn your behaviors,

© 2016 Phocuswright Inc. All Rights Reserved. 6 Travel Chatbots September 2016

respond to stimuli and react in unpredictable ways. Other ways of using AI include assisting in fraud detection, online customer support as well as movie and music recommendation services.7 Unfortunately, expecting a fully automated, machine­driven chatbot to provide all the services requested by the traveler is not realistic at this point of time. The firestorm created by Microsoft's AI­driven chatbot Tay, that began posting racist and sexist messages on Twitter in response to questions from users, is an example of the current limitations of AI­driven bots.8 The common opinion voiced by the majority of travel industry presenters at the MobileBeat conference is that a blend of AI­driven bots combined with human intervention is the best practice today. In fact, there are a number of services that do not deploy a chatbot at all, but simply rely on human text conversations.

Are Chatbots an Interim Step to Voice­Enabled, Human­to­Machine Conversations?

During the MobileBeat conversation, Kathleen McMahon, vice president and general manager at SoundHound, presented their new platform for voice­enabled automated conversations. She argued that the current chatbot craze was merely an interim step that will fade once real voice­enabled human­machine conversations become the norm. However, it is more likely that both chat­based and voice­based conversational interfaces will co­exist and be used based on the preference of the individual. Millennial and younger generations, whose life is centered around text­based chat, may prefer chatbots to voice. Ultimately, whether voice will replace chat will be driven by customer preferences and the accuracy and simplicity of the specific solution.

Chatbots and the Impact on Search­Shop­Buy

The question facing the travel industry is to determine the most effective use of chatbots within the Search­Shop­Buy process. The mainstream online travel companies who participated in the MobileBeat 2016 conference uniformly voiced their concerns over their ability to completely replace human interaction. In fact, a clear warning emerged from the speakers and panelists, who included industry heavyweights such as , Booking.com, Airbnb, Trip Advisor and Uber, that a bad experience with a chatbot could negatively impact the overall travel brand. In general, all speakers and panelists at the conference emphasized the fact that chatbots for commerce is at a very early stage of development.

The use of chatbots may have an impact on all stages of travel Search­Shop­Buy (e.g., Dream­Search­Shop­Buy­ Experience­Share). Below are a few user case scenarios where chatbots may play a role.

■ Leisure travelers without a specific vacation destination (Dream): A leisure traveler might aska chatbot to "show me beach vacations" without specifying a particular destination. The challenge with this example is having enough content to meet all the possible options for the traveler in a concise­enough way within the constraints of chat. The value of the response also depends on the chatbot understanding the preferences and personas of the traveler. Without a basic level of personal insight, the chatbot is reduced to simply suggesting a beach destination that is popular with other users, potentially ignoring the traveler's true preferences. The traveler also may be forced into an extensive dialogue to state their explicit preferences, limiting the efficiency of the chatbot conversation. ■ A business traveler looking for a specific hotel in a city (Search): For a simple request such as a limited search for a company­approved hotel near a client or company location, a chatbot could provide a quick response that is policy compliant. This requires back­end connections to corporate policy systems and preferred vendors within a managed travel context. As with the leisure example, the traveler's preferences would need to be known to make the process efficient.

© 2016 Phocuswright Inc. All Rights Reserved. 7 Travel Chatbots September 2016

■ A leisure or business traveler comparing similar hotels via a chatbot (Shop): If a traveler has already narrowed down their search to a limited number of lodging options, a chatbot could be used to compare similar properties based on amenities or reviews. For example, "for the three hotels being considered, show me which has the highest customer rating. "As in the previous example, the value of this shopping activity is dependent of the breadth of back­end connectivity and content and the ability to shop on multiple parameters. ■ A frequent user initiates a seamless purchase via a chatbot (Buy): The ability to execute a seamless chat­based purchase is dependent on the amount of information already provided to the service. An existing OTA, TMC or metasearch company may have payment information already stored, making a simple a "buy" message possible. In this type of scenario, once the buy request is submitted to the chatbot, the stored payment information is used to complete the transaction. ■ A guest requesting a WiFi password and dinner recommendations at a hotel or resort via a chatbot (Experience): The ability to get advice from a chatbot in destinations holds a great deal of promise for the conversational technology, especially if the requests are routine. The chatbot could respond to simple requests such as a WiFi password, eliminating the need to talk to a hotel staff member. A more sophisticated chatbot could combine location, destination­specific information and personal preferences to recommend a restaurant, tour, activity or local merchant. Sophisticated chatbots can identify and classify common service issues and could help solve problems or filter requests to allow humans to solve problems more effectively. The more complex scenario would need to gather sufficient personal preferences to be an effective channel for local activities and still may require human intervention to complete the transaction. ■ A family returns from a resort vacation and shares their feedback (Share): Often travelers only answer trip surveys if they are unhappy or very satisfied. Chatbots could play a role in encouraging candid feedback. With enough insight around the customer's actual experience, chatbots could ask specific targeted questions, such as whether the children enjoyed the resort's day camp or if the father liked the golf course. Chatbots used in this fashion could help travel providers pinpoint feedback and personalize the often impersonal post­trip survey.

Examples of Travel Bots and Human­Driven Texting Solutions

Text­based reservation platforms have been growing within the travel sector. In some cases, there is no chatbot deployed and you are simply communicating with a human agent, but even these human­based chat sessions may deploy AI behind the scenes to improve response.

Below is a list of both chatbot and human­assisted, messaging­based travel solutions. This list is subject to change as new entrants constantly emerge and others fade or are acquired, so please forgive any inaccuracies or omissions.

New Leisure Travel Services

■ Chatnbook (www.chatnbook.com/) — a chatbot promoting direct hotel reservation and collaborative travel planning.

© 2016 Phocuswright Inc. All Rights Reserved. 8 Travel Chatbots September 2016

■ GoButler (http://www.angel.ai/) — originally launched as a messaging based general virtual assistant. In March, it relaunched with a focus on travel. More recently, GoButler has pivoted again, no longer offering a B to C app but rather a platform for Natural Language search for commerce under the Angel.ai brand, still using flight search as demonstration of their technology. ■ GoHero (http://www.gohero.ai/#six) — a travel chatbot that uses NLP to process customer requests and is supported by human staff. The requests are taken care of by the backend booking engine which uses multiple APIs. ■ Hello (http://helloscout.com/home/) — a team of in­destination concierge manned by local experts. It currently does not use chatbot technology. ■ KimKim (https://www.kimkim.com/) — a travel service that connects travelers with local tour and activity providers. KimKim recently launched a chatbot to gather upfront information. ■ Lola (https://www.lolatravel.com/) — it currently provides human chat with travel agents using AI in the backend to assist the agent in providing customized travel planning/booking services. ■ Mezi (https://mezi.com/) — a personal shopper and end­to­end travel assistant that facilitates booking flights, hotels, restaurants and entertainment options, as well as taking care of cancellations or changes to plans while en­route. The service uses a combination of AI and human travel agents. ■ Pana (https://pana.com/) — a human­based travel agent service, the company states that some advanced technology is used behind the scenes to facilitate the human agent and recently introduced a chatbot.9 ■ Skylark (https://skylark.com/#/?_k=tr1iwo) — a human­based travel assistant targeting luxury travel. The service currently does not use chatbot technology. ■ Snap Travel (https://www.getsnaptravel.com/) —it uses a chatbot to automate hotel reservations. The chatbot provides automated responses and assistance from backend support staff.

OTA

■ Booking.com (http://www.booking.com) — The OTA has implemented two types of chat: one allows travelers to communicate directly with their hotel about aspects of their stay. This is not using chatbot technology.10 More recently, Booking.com launched a pilot in Amsterdam that leverages machine learning and artificial intelligence to predict users' travel intent and provides customized local activities and tours with streamlined payment options and priority queueing using chatbot technology.11 ■ Expedia (https://www.expedia.com) — In June, Expedia launched a chatbot for Facebook Messenger designed to helps travelers book hotels.12 The bot uses NLP and Machine Learning to create a structured conversation flow. It analyzes information provided and prompts the user to input other relevant data points to complete a search. 13 © 2016 Phocuswright Inc. All Rights Reserved. 9 Travel Chatbots September 2016

■ HotelTonight (https://www.hoteltonight.com) — The mobile­only, last­minute hotel booking app introduced Aces, a concierge service via chat to help customers with their hotel experience and stay. The service is powered by humans (local "aces" in each destination). ■ Ixigo (https://www.ixigo.com/) — The Indian OTA launched an AI­powered chatbot to help the travelers plan their trip by providing information on flights, hotels, buses, cabs, places to visit, vacation destinations and more. Travelers have the option to either use Ixigo's proprietary meta­search engine, or the chatbots find the cheapest travel deals.14

Corporate/Business Travel Services

■ ETA (http://www.eta.ai/) — Still in stealth mode, this new message­based AI assistant creates trips based on calendar entries and helps the user manage their entire journey. ■ Hyper (https://www.usehyper.com/) — Originally launched a chatbot for travel, but was recently acquired by Tradeshift and is now part of an overall intelligent assistant effort. ■ 30 Seconds to Fly (https://www.30secondstofly.com/) — The Company's CLAIRE is a travel chatbot designed to bring travel management capabilities to the Small Medium Enterprise (SME) market. It utilizes NLP and Machine Learning to help direct travelers to options within corporate policy.

Hospitality

■ Checkmate (http://www.checkmate.io/) — (now owned by TrustYou) An integrated messaging platform for hoteliers. No current deployment of chatbot technology. ■ GoMoment (http://www.gomoment.com/) — A hotel chatbot that is designed to automatically answer hotel guest questions. Also proactively offers mobile checkout and other services. ■ GuestDriven (http://www.guestdriven.com/) — Provides a messaging for guests and hotel staff and allows personalization of the guest stay. No current use of chatbots. ■ GuestU (http://guestu.com/) — A mobile concierge for hoteliers. The platform app uses a chatbot for basic hotel questions and provides a messaging capability for the hotel staff.

■ Hyatt (https://www.hyatt.com/) — The hotel operator has Implementedhuman­based chat available on Facebook Messenger, but no current use of chatbot technology.15 Hyatt has selected Zingle as its preferred guest messaging service for Hyatt hotels worldwide.16

■ IHG (https://www.ihg.com/) — IHG has developed a chatbot/virtual assistant for Facebook Messenger.17 ■ Kipsu (https://www.kipsu.com/) — The start­up provides a messaging platform to multiple industries. One target market is hospitality­facilitating text messaging between guests and staff. No current use of chatbots. ■ Marriott (http://mobileapp.marriott.com/mobile­requests/) — In 2015, the hotel brand launched Mobile Request. Members who have upcoming reservations are able to communicate with their hotel in two ways. The "Anything Else?" feature offers guests two­

© 2016 Phocuswright Inc. All Rights Reserved. 10 Travel Chatbots September 2016

way chat functionality to have conversations in real time with the hosts at Marriott Hotels, who can fulfill and confirm their requests. It also offers a drop­down menu with most requested services and amenities, such as extra towels and pillows.18 No current use of chatbot technology. ■ Moncierge (https://www.monscierge.com/) — A concierge and message platform for the hotel industry.No current use of chatbot technology. ■ Radisson Blu Edwardian Hotels, London (http://www.edwardian.com/) — The company created a chatbot virtual host, Edward, that is available at 12 properties across London and enables guests to request amenities such as fresh towels or room service via text message.19 ■ Whistle (http://www.trywhistle.com/) — The service provides a messaging capability on an existing messaging platform for hotels facilitating text messaging between guests and staff. No current use of chatbot technology. ■ Zingle (https://www.zingle.me/) — The software providesa messaging platform to multiple industries. One target market is hoteliers facilitating text messaging between guests and staff. No current use of chatbot technology.

Metasearch

■ Hipmunk — launched an AI­powered chatbot assistant that provides automated advice to would­be travelers. ■ — launched an interactive chatbot for Facebook Messenger that enables search and booking of travel inside the app. The new tool lets users find flights and hotels using conversational language.20 The metasearch company has also enabled voice search on Amazon's Echo platform. ■ ­ launched a bot for Facebook Messenger that facilitates the search and booking of flights.21 The metasearch company has also enabled voice search on Amazon's Echo platform.

Airline

■ Icelandair — launched a messaging bot on Facebook Messenger that drives users to book stopover flights in Iceland while visiting other overseas countries.22 ■ KLM (https://messenger.klm.com/ — the first airline to launch a chatbot on Facebook Messenger that provides an automated delivery of documents, status updates and a conversation path for customer service issues. ■ Virtual Agents — Over the last 10 years, many airlines have launched chatbot virtual agents created by companies such as Intelliresponse (acquired by [24]7 in 2014 (http://www.247­inc.com/), Next IT, (http://www.nextit.com/), Artificial Solutions (http://www.artificial­solutions.com/) and Inbenta (https://www.inbenta.com/en). 23 There are numerous examples of virtual agent bots on airline websites including Alex (United), Sofia (TAP Portugal), Virtual Assistant (Iberia) and Ana (Copa). Some of the virtual

© 2016 Phocuswright Inc. All Rights Reserved. 11 Travel Chatbots September 2016

assistants use the more traditional SLT approach to chat, while others have embraced NLP and Machine Learning. The positioning of these virtual agents as a standalone website assistant is quite different from the current chatbot wave that runs on top of Facebook Messenger or Slack.

Ground Transportation

■ Uber — has rolled out a chatbot on Facebook Messenger that allows users to order cars from any Messenger conversation by tapping a car icon. The user can also message directly with Uber to order a car, but it requires the user be logged into their Uber account through Messenger for this feature to work.24

Summary

Have we reached a stage where chatbots for travel are ready to replace human beings? The answer is still both yes and no. Even with the advancement of NLU and Machine Learning­based chatbots, a fully automated text­based travel reservation conversation is only effective if the chatbot fully understands the query. The chatbots are smarter, but not quite smart enough to replace a human conversation. The travel industry speakers at MobileBeat uniformly voiced this belief and warned the audience that a poor interaction with a chatbot representing your brand is a dangerous experiment. That being said, it is very clear that chatbots can play an important role in customer service and capturing initial travel search queries. Although humans remain very much in the mix for now, with the rapid advancement of AI, a true intelligent chatbot capable of complex conversations is on the horizon for the travel industry.

© 2016 Phocuswright Inc. All Rights Reserved. 12 Travel Chatbots September 2016

Endnotes

1. Jared Newman, How The New, Improved Chatbots Rewrite 50 Years Of Bot History, Fast Company (May 2, 2016).

2. Justin DiPietro, The human role in a bot­dominated future, Tech Crunch (July 30, 2016).

3. Jonathan Batson, Forget the Turing Test: Here's How We Could Actually Measure AI, Wired magazine (June 12, 2014).

4. Marco Luca Sbodio, From knowledge graphs to cognitive computing, IBM Research Blog (Jan. 18, 2016).

5. Gavin O'Malley, Bots Are Taking Over Facebook!, MobileInsider (July 21, 2016).

6. Josh Constine, Slack's growth is insane, with daily user count up 3.5X in a year, Tech Crunch (April 1, 2016).

7. "10 Examples of Artificial Intelligence You're Using in Daily Life," Beebom (Dec. 1, 2015).

8. Mathew Ingram, Microsoft's Chat Bot Was Fun for Awhile, Until it Turned into a Racist, Fortune Magazine (March 24, 2016).

9. P. Pragati, CHATBOT | #Human Machine #Artificial Intelligence, TELCO Professionals (July 27, 2016).

10. Matt Marshall, Booking.com launches a chat tool to connect hotels and travelers, Venture Beat (May 3, 2016).

11. Alex Samuely, Booking.com anticipates mobile users' purchase intent with AI­driven experience, Mobile Commerce Daily (July 15, 2016).

12. Zerega Blaise, Expedia's first bot is for booking hotels, Venture Beat (June 8, 2016).

13. Ibid

14. Ixigo Launches AI Powered Chatbot Ixibaba, Inc42 (July 21, 2016).

15. Are new chat bot platforms right for hotels?, Hotel News Now (May 11, 2016).

16. 06 APR ZINGLE TEAMS WITH HYATT AS GLOBAL GUEST MESSAGING SERVICE, Zingle press release (April 6, 2016).

17. Nancy Trejos, IHG to start communicating with guests through Facebook Messenger, USA Today (July 1, 2016).

18. John Wolf, Travelers Can Ask Anything, Anytime, Anywhere with the New, Industry­leading Mobile Request Chat Feature from Marriott Hotels ­ Now Available on the Marriott Mobile App, Marriott press release (May 13, 2015).

© 2016 Phocuswright Inc. All Rights Reserved. 13 Travel Chatbots September 2016

19. Sophie Witts, Edwardian Hotels London launches robot concierge, Big Hospitality (May 10, 2016).

20. Paul Brady, Kayak's Facebook Messenger Chatbot Makes Trip Planning Even Easier, Conde Nast Traveler (June 29, 2016.

21. How to search for flights with Skyscanner's new Facebook Messenger bot, Skyscanner News (May 26, 2016).

22. Brielle Jaekel, Icelandair ushers in stopover bookings via chatbot on Facebook Messenger, Mobile Commerce Daily (Aug. 4, 2016)

23. Chatbots.org https://www.chatbots.org

24. Tess Townsend, 5 Chatbots You Can Try Now on Facebook's Messenger, Inc. Magazine (April 15, 2016).

© 2016 Phocuswright Inc. All Rights Reserved. 14