Entropia study claims chatbot experience hindered when only developed on Facebook

Chatbots are slowly becoming commonplace in this day and age as more brands jump on the bandwagon to create more efficient and personalised user experiences. While it is common for brands to use chatbots as a customer representative, most of them are built on Facebook's Messenger platform.

According to Entropia's recent study titled "Bot-o-logy" which examined 30 global and local chatbots across reliability, experience and personality, companies seem to place "reduced emphasis" on their chatbot's user experience by building it on Facebook Messenger. The Messenger platform offers a simple chatbot engine that brands and individuals can use to create their bot within hours. Coupled with Facebook's proprietary AI, brands can create a basic chatbot with little to zero knowledge in that area.

Although these basic chatbots can be created at a fraction of the development cost compared to creating it from scratch, developers will be limited in terms of what they can do to create a truly seamless chatbot experience.

Chatbots that excel in the experience criteria group tend to be from brands that went the extra mile to build their chatbots on reputable platforms outside of Facebook that offer more options for customisation, Entropia's study found. It added that these chatbots tend to learn and react much faster, as the needs of the brand are taken into account without compromise.

Entropia's study listed three types of chatbots. Rule-based chatbots comprise conversations that are prefixed and users are unable to enter text outside of a predefined script. Natural language processing (NLP) based chatbots are conversational but can only understand natural language by learning via preset content. Lastly, AI-based chatbots learn through conversations and automatically build responses.

In terms of experience, the chatbots scored the lowest in accessibility and cross devices as the Messenger platform is unable to exist outside the Facebook environment. As such, users are required to log in to Facebook to initiative a chat, and both brands and developers will have to rely solely on Facebook to function across more devices.

That said, although the Messenger platform offers less options for customisation and creates limitations for developers, it also has advantages such as allowing developers to focus more on populating the chatbot with content without worrying about technical issues such as load time, refresh rate and database crawling. The study added that the Messenger platform is already well-optimised, making crawling through and pulling information much faster, compared to developing a chatbot from an original platform.

Also, the chatbots scored overwhelmingly well on quick selections and ease of messaging. The study said this comes as no surprise as most chatbots today are built on established chat platforms such as Facebook's Messenger. The standard user interface makes it easy to navigate for most users who are already familiar with the format. Despite being user-friendly,

Meanwhile, companies also focus on reliability in their chatbots so that conversations with consumers can come to a productive end. Most chatbots that score well in this area have "concrete amounts" of information in their database, which lets them provide more relevant responses to user queries. By integrating natural language with a wealth of user information, chatbots can handle conversations that escalate outside their territory of expertise and respond to users who are there just to chat.

On the reliability front, Entropia's sampling of chatbots scored highest on ease of navigation and response time. The study said that this is an indicator that most brands place more emphasis on certain aspects of the customer journey. With predefined user responses, these chatbots can walk users through to a goal state set by the developer. However, guided chats can never build up to support natural conversation, nor handle new and non-linear user queries. In the meantime, conversation depth scored the lowest across the board for the reliability criteria. Only a few chatbots have managed to handle and predict the many directions conversations can go.

"In this case, Facebook's Messenger platform is both a blessing and a curse - developers are forced to compromise on integration with the company's local environment (and the depth of their chatbot interactions) to minimise development cost and time spent," Entropia said.

Brand relevance scored the highest when it came to the personality criteria. With chatbots being one of the first consumer touch-points when it comes to interacting with a brand, Entropia said it is  encouraging to see brands making sure their relevance shines through, even in conversations. While the idea of having an AI chatbot as an employee can be new to brands, it is possible to inject a brand's employee criteria into AI-based chatbots.

Whether or not the chatbot works well or provides enough information, brand essence should be present in the conversation, the study said. So far, brands have not managed to crack localisation, which has the lowest scores for personality. Meanwhile, many brands have opted to use just Standard English for their chatbots due to difficulties in implementing local languages and slang.

[Digital Marketing Asia Conference 2019 in Singapore is back! Join us on 8-9 October as we hear from experienced practitioners and thought-leaders on how they are managing complex digital transitions and reimagining new ways for their marketing to become more customer focused, agile and interactive. Book your seats today.]

What makes a great chatbot?

The highest rated chatbot, according to the report, is Mitsuku, a virtual character with a unique personality developed by AIML Technology in Leeds, England. This chatbot is the most notable high scorer across the criteria of reliability, experience and personality as its responses are always speedy and factually correct. The lowest scorer, according to the report, is Endurance chatbot, developed primarily to be a virtual companion to the elderly, that performed poorly across most parameters. Leading the local pack with high scores on many criteria, is AirAsia’s Virtual Allstar, which is a good example of how, through integration, a chatbot can provide a seamless experience to its customers in a quick and efficient way.

According to Entropia, great chatbots go beyond frequently answered questions (FAQ), make themselves available, are continuously improving, and represent their brand well.

Too many chatbots today exist purely as FAQ machines. Plenty of users prefer to bypass automated responses to get to a human customer representative because they feel like they can ask more specific and personalised questions. The study said that chatbots should be prepared to interact and respond to users naturally, beyond scripted queries, to satisfy a broader range of user queries.

Additionally, great chatbots connect with users the way their friends do. Making a chatbot available on the users' most preferred platforms will naturally facilitate more interactions and queries. Chatbots that incorporate new accessibility technologies such as voice-to-text and image recognition will also be able to enhance the way brands and customers communicate.

The study added that brands should consider chatbots an ongoing investment. A chatbot is only as smart as the information it has access to, yet the study found that many chatbots are not even integrated into the
brand's information database. Chatbots should be regularly updated to include new customer information; and improvements should be made to the way it handles natural language queries.

Although most brands know enough to get the tone of voice right, chatbots today need to be integrated into the company's ecosystem to provide personalised responses according to individual needs and circumstances, minimising redirection to human representatives and external sites.

Prashant Kumar, Entropia founder and senior partner, said a vast majority of bot interactions in Malaysia tend to be "deeply frustrating". This brings the fear of the very technology getting discredited.

Sourabh Agrawal, partner at Entropia added that on the home front, although chatbot technology is still relatively new to Malaysia, progressive companies, always eager to innovate and automate, have adopted these online conversation agents as an efficient alternative to improve their customer communications.

"And due to the ever-increasing capabilities of Artificial Intelligence, there is no limit to what chatbots can do today. In fact, there is huge potential for future-focused brands to carefully design and develop their online conversation agents to match different market segments," Agrawal added.

This research by Entropia claims to provide an analysis of some of the best and worst chatbots in the market today, including those of local brands like Air Asia, Malaysia Airlines, Digi, as well as select international brands like Ford, Shell, Wall Street Journal, Star Wars, Western Union, TGI Fridays, Endurance, and Mitsuko.

(Photo courtesy: 123RF)