PwC strengthens its disruptive digital technology credibility. Starting today, over 1.5 million insured persons and interested parties can experience our expertise first-hand: the major health insurer SWICA has successfully launched ‘IQ’, the Swiss health insurance industry’s first artificial intelligence (AI) chatbot. The natural language processing bot has been developed by SWICA in collaboration with the software provider Enterprise Bot and the people from PwC’s advisory team.
Chatbots: opportunities and challenges
Chatbots like IQ are innovative digital programs that conduct conversations using textual methods; they are transforming forms of customer engagement as they meet customers’ demand for an enhanced user experience through AI. As customers seeking support in everyday decision-making have a preference for using mobile solutions, chatbots are generally integrated into messaging platforms – as this is where customers want to interact with their friends and businesses alike (PwC, 2017).
Specifically, in interactions with customers, chatbots can deal with routine queries by providing smart, omni-channel support while delivering consistent responses and 24/7 individual advice (e.g. regarding customer service and policy management). This service increases customer satisfaction and frees up human staff to handle complex requests (PwC, 2018a).
In addition, within the organisation, the operation and maintenance (e.g. content management) of chatbots will connect your people and departments more effectively (i.e. customer service, claims management, marketing, IT and digital services) and improve your work efficiency, knowledge management, processes and services.
Natural Language Processing (NLP) – the heart of the AI chatbot ‘IQ’
IQ is based on NLP, which uses methods and techniques from linguistics combined with modern computer science and AI. The aim is to allow direct communication between humans and computers based on natural language (e.g. open questions). For this, the users’ language is processed by applying algorithms, procedures of AI and machine learning.
A challenge for NLP-based chatbots is the complexity of the human language, for instance in linguistic ambiguity and the presence of various dialects. That is why big data techniques are very important for AI chatbots like IQ: in order to recognise text meanings holistically, it is necessary to capture large amounts of data in advance (‘train the bot’).
“IQ is more than a tool to answer simple questions. The chatbot is another piece in the puzzle of digitisation and round-the-clock accessibility.”
Reto Dahinden, CEO SWICA Healthcare Organisation
Everything started with a proof of concept (PoC)
In 2017, SWICA developed a PoC chatbot together with PwC and Enterprise Bot to test whether a chatbot could achieve the necessary response accuracy and customer friendliness required for the demanding and complex Swiss health insurance market. SWICA’s main challenge at that time was the high burden on customer service (incl. stand-by service) due to repetitive standard questions related to SWICA’s large portfolio of insurance products. Using an AI chatbot was expected to free up customer agents to be able to focus on complex requests and to be a starting point for the improvement of omni-channel management.
The team from Enterprise Bot and PwC were able to live up to these expectations: after the successful PoC phase in the summer of 2018 (PwC, 2018b), which included intensive and positive testing sessions with customers and the SWICA departments, the project continued in spring 2019. The functionalities of IQ were refined and its knowledge expanded: IQ can now provide information and recommendations on SWICA’s insurance products and respond to questions regarding the Swiss healthcare system (KVG, VVG and UVG) and SWICA as a company.
During SWICA's internal testing phase, more than 90% of the conversations were processed and answered independently by IQ. This will greatly relieve customer services, whose on-call service is currently available 24/7. For now, IQ answers questions in German only. The chatbot’s progress will be carefully monitored by SWICA. The French, Italian and English versions of IQ are expected to follow by the end of 2019.
Above all, the success of the project is the result of great teamwork within the following project team:
Feel free to contact us for a more in-depth chat about how to boost your business with AI chatbots.