Publication: Asia Insurance Review
In recent years, chatbots have graduated from being vanity projects in many insurers’ corporate innovation labs to real-world use ranging from customer service and consumer marketing to agent recruitment and frontline engagement. This evolution is especially prevalent in the insurance sector in Southeast Asia. Pand.ai's Shin-Wee Chuang provides some insights.
Most of the insurance chatbots in this region today are text-based, instead of voice-activated. In Singapore, Tokio Marine launched the industry's first AI-powered chatbot, TOMI, in January 2017. A few months later, this was followed by other insurance chatbots such as AskPru from Prudential, Jiffy Jane and RevoRetire from NTUC Income, Mae from MSIG and Sompony from Sompo.
In Malaysia, AIA launched AskSara for its agents in October 2017, the first insurance chatbot in the country. A year later, Allianz launched AIDA, the first consumer-facing chatbot in Malaysia promoting its Smart Home Cover home insurance product.
What can a chatbot do?
At its core, a chatbot is essentially a platform that matches user input, or questions, to a pre-defined database of intents to retrieve a response. A basic chatbot, even if AI-powered, usually fetches responses from a text-based database for users.
A more advanced chatbot, on the other hand, is usually infused with advanced software engineering and design strategies to perform more complex tasks. These would include user-specific enquiries like claim status and premium due date, transactional based interactions like claim submission and real-time report generation, and handover mechanisms such as live chat integration.
However, it is important not to confuse a chatbot's ability to 'match' a question to an answer to its ability to 'generate' an answer. Natural language generation (NLG) is a software process that transforms data into written narrative. While NLG could, in theory, generate an answer from raw data, this technology is currently in its infancy and is not expected to be commercially read for some years. Until then, it is best not to dump uncleaned, raw data into the chatbot and expect it instantly to produce an answer that is not only accurate, but also meets compliance guidelines.
Do chatbots really work for insurance?
Insurance companies are early adopters of chatbots in Asia. This is perhaps because traditionally, the insurance sector has always been a business that requires considerable facetime between the customers and its agents.
Chatbots act as a natural extension to the digital world while preserving the human touch through conversational AI. There is still much debate on whether chatbots really work for the insurance industry, partly due to a lack of consensus on what constitutes a successful chatbot implementation.
Some insurers insist on using 'accuracy' for measuring the success of a chatbot, while others alll back on business KPIs such as sales and revenue. The most enduring chatbots, however, are the ones that evolve and improve over time, such as TOMI from Tokio Marine Life Insurance Singapore (TMLS).
Case study: The evolution of TOMI
TOMI was born in January 2017 and went through two main evolutions alone in its first year of operation. It started as an automated help desk for TMLS's agents, then was quickly extended to frontline engagement with both agents and financial advisers before surfacing as a public-facing chatbot in October 2017.
The first phase was exclusively for TOMI to interact with insurance agents. TOMI was launched shortly before Chinese New Year and included an in-built gamification module that allowed TMLS's agents to win an 'Ang Bao' (a traditional Chinese custom of red packets containing money) each time an agent correctly answered all the questions posed by the bot.
TMLS used this seasonal campaign to draw in agents to use TOMI and achieved about an 80% take-up rate. The number of users who interacted with TOMI was consistently at 98% in the first three months. TMLS later used TOMI not just as a help desk, but also as an engagement tool to push relevant articles, news or tools to the agents. In April 2017, TOMI was extended to financial advisers in the TMLS network.
In October 2017, TOMI was officially launched to the public on Facebook Messenger. Within three months it garnered more followers than its Facebook page had in three years, without any marketing support. Since then, TOMI has been used mainly as a test-bed for new ideas and innovations. This includes agent recruitment where TOMI successfully combined online registration with offline events via chatbots. The visitor-to-registration ratio for these recruitment campaigns was consistently above 25% and achieved 55% at its peak.
Leads generation is another area where TOMI has made a considerable impact. TOMI has achieved a 10x visitor-to-lead conversion rate compared to more traditional conversion channels or leads conversion through static websites. TOMI's latest product marketing campaign received an unusually high 50% visitor-to-lead conversion. This was possible through the usage of myriad proprietary conversation tools, such as the bot's ability to understand a complex request that has various parameters in a single text question, and modules such as personalised reports on its platform.
TOMI also shows promise in customer services. Although not directly integrated with TMLS' core systems yet, TOMI is able to cater to clients' preferred choices and provide efficient customer service. For example, a client requesting the approved panel clinic would be able to ask questions such as 'Find me a female doctor for X-ray in Bishan'. TOMI will then provide an answer that matches those exact requirements.
Using TOMI as a testbed for innovation and ideas is a strategy that has paid off for TMLS. This was made possible because of the team's focus on using actual usage data to guide its product's roadmap. The team at TMLS expects the core natural language processing (NLP) technology to keep improving, but the biggest win for TOMI could come from improved user experience through personalization.
The AI/NLP technology used to power chatbots has yet to mature. However, it does not mean that chatbots cannot be useful commercially. Of course, for every successful chatbot, there are unsuccessful ones that did not make the cut, which is only to be expected because of Asia's pioneering role in chatbot adoption.
TOMI's successes demonstrate that chatbots can work for insurance. While still in the early stages of adoption, there is a lot of potential for chatbots to improve and change the insurance landscape in time to come.
Mr Shin-Wee Chuang, is the co-founder and CEO of Pand.ai.
Learn more at www.pand.ai
Request a demo at www.pand.ai/contact