Chatbots are the new face of customer service. A study by Forrester in 2018 found that 57% of companies globally already using chatbots or planning to do so and the chatbot market is surely picking up speed in Asia Pacific, according to Global Market Insights. It is said to generate revenues amounting to approximately US$350 million by 2024 due to the advent of messaging applications such as WeChat in China.
As such, customer service as we know it has changed – driven by customers’ increasing priority on quick responses. But at Marketing Interactive‘s Digital Marketing Asia 2019, Felix Yuwono Kurniawan (pictured below), head of product – chatbot, dispute settlement and operation at Tokopedia said speed alone is no longer suffice to meet customers’ expectations today.
Tokopedia started off with a chatbot that replies to frequently asked questions automatically in early 2018, but Kurniawan said consumers were not satisfied with the answers it provided. The chatbot, which was tested among 10 users for three months, “failed miserably” as nobody used it and it took Tokopedia six months to roll out a chatbot that is good enough for the general public.
According to Kurniawan, people are looking for a chatbot that is conversational, and communicates more like how a human would. “People are changing. They are expecting more and more. They want the chatbot to be more human-like, and they want a good user interface, artificial intelligence (AI) and system integration,” explained Kurniawan. The chatbot should also be user-friendly, so customers know how to use without any onboarding.
“It is important for businesses to get these aspects right, or consumers may lose trust in the chatbot and not use it anymore,” he added. The chatbot also has to adopt a suitable tone in its communication to align with the brand and protect the brand reputation. He said:
You have to create good responses as it reflects the persona of your brand.
To ensure that the chatbot is able to cover types of queries, spanning delivery, product, payment and login, Tokopedia invested in system integration. Doing so enables the chatbot to provide more exact answers, rather than generic ones. It also regularly checks for “fall backs”, where the chatbot is not able to answer and trains it to solve them.
The data of the chatbot is constantly being refined to provide better auto-fill suggestions and reduce “false positives” where mismatched answers are given. The team at Tokopedia makes sure to review the cases where customers decide to seek a human agent instead or gives their chatbot experience a bad rating.
Today, Tokopedia’s chatbot is able to make small talks, fulfill routine requests as well as up-sell products personalised based on queries. It currently completes 55% of Tokopedia’s enquiries and sees more than 65% user engagement per month, and is now the first point of contact on its mobile app, followed by live chats. It has also increased efficiency, saving the company millions of dollars, said Kurniawan. Since the Indonesian eCommerce company embarked on its chatbot journey two years ago, it has grown its team of software engineers and data scientists from two to 10. AI trainers have also more than doubled to 20. Tokopedia is currently putting a platform as a service to help other companies to build their chatbots.
Benefits of chatbots
Also speaking at the event was Evan Tanotogono, head of digital channel at Asuransi Jiwa Sequis Life (Sequis), who said that the removal of human interaction in chatbots could be a plus for customers. He explained that in his industry, there is a subtle negative sentiment towards life insurance in Indonesia due to the way it is sold. As there are very limited insurance agents outside Java, Indonesians get a lot of calls from telemarketers and they do not like in.
Market research by Sequis has found that one of the pain points customers faced is that they do not feel like they are in control with the telemarketers. They want to have alternative browsing channels to see the full suite of offerings. In light of this, Sequis revamped its website and transactional flow to position itself as a digital insurance brand.
But having a full digital platform is not enough. Tanotogono said: “It takes away the luxury of having personal chats, of people interacting and asking questions, and the warmth of the brand.” He explained that insurance is a complex product that needs explanation, cross-sell and up-sell – essentially, it requires conversation.
This led the company to recently launch a chatbot, which is currently in soft-launch mode. Chatbots can be a great educational tool but Tanotogono said in an industry that is heavily regulated, marketers have to put in place the necessary safeguards. For example, the chatbot has to be able to recognise inaccurate inputs such as typing errors in fields that require critical information. If the user is not eligible for an application, he or she should also be disallowed to proceed.
Why brands can’t do without chatbots
SEA surpasses US and India in adoption of conversational commerce, finds FB study
Why Mah Sing’s AI-enabled Madam Loan bot is the antithesis of a typical chatbot
Entropia study claims chatbot experience hindered when only developed on Facebook
Are your bosses really buying into CX?
Why emerging tech opportunities continue to slip through fingers of marketers
‘Our chatbot was by no means the first in market. That was really deliberate,’ says AirAsia