The Chatbot Prototype is designed to increase the amount of time a user spends in app, allowing the client to target growth of particular products through personalised recommendations for users, and allowing the client to learn more about the user through their interactions with the bot.
The Chatbot Prototype was also designed to provide the long-term, ongoing benefit of reducing cost to serve by decreasing the load on customer service centres.
Chatbot technology is an emerging area that has seen strong interest from major technology companies such as Microsoft, Facebook, and Google. As an expression of current trends in AI and Machine Learning, Chatbots represent a new and innovative way for businesses to engage with mobile users, learn more about user behaviour, and market their products through targeted and personalised recommendations.
By nature, chatbots rely less on design appearance and more on the text copy itself to create personality. The visual design of the Prototype is clean and simple, to ensure that nothing detracts from the chat experience. On-boarding was kept to a minimum, with a focus instead on establishing interaction between the user and the Chatbot, and utilising the chat to obtain the necessary information from the user. The use of guided conversation required that the pre-filled responses at the bottom of the screen be recognisable to the user as tappable buttons, which influenced the design choices made for these elements.
The Chatbot Prototype explores major existing machine learning technologies and platforms, such as Amazon AWS Machine Learning, IBM Watson and Google TensorFlow to develop a solution compatible with the client’s objectives. Using data mining techniques, the Chatbot gathers data from the public domain to determine trending topics of interest, and uses existing data from user interactions to continually refine the user experience. Machine learning technologies allow the chatbot system to be trained over time to provide users with more targeted conversations based on their previous interactions with the system. The solution has the potential to incorporate an unlimited amount of data sources to provide a more intuitive and targeted experience for users, and the features of the prototype can be altered to suit the client’s industry and objectives.
Key features of the Prototype:
- Reduces cost to serve by handling user enquiries
- Ability to learn from user habits to find out topics of interest
- Uses questions to determine a selling point
- Ability to fetch real-time information such as exchange and interest rates
- Option to link user to a map or contact card to locate nearest store
- Option to link user to a customer service representative within the same chat if the user has more complex enquiries
- Ability to learn from pre-existing chat data to improve user experience
- Potential to learn from the user’s other interactions in the app to improve targeted services
A key learning from the project was the difference between open and guided dialogues.
In an open dialogue, the user can ask anything, and the chatbot will learn from its conversations with users. However this technology is still in its early stages of development, and the open-ended structure of the conversation means users don’t always know what to ask the chatbot.
In a guided or closed dialogue, the user is presented with a set of predetermined responses to the chatbot’s dialogue. This mitigates the risk of misunderstanding within the chat, and creates a better user experience. This technology has already been successfully applied elsewhere, in applications such as Quartz, Lark and Penny.
Possible applications of this prototype include:
- Banking and Financial Services
- Customer service functions within native apps
Other possibilities to consider for future development include:
- Voice integrated chat interaction