China’s biggest banks have embraced chatbot technology to serve customers far beyond what peers are doing in other markets.
“China mainland’s [automated] banking customer service is far ahead of Hong Kong or United States,” said Chen Shujin, director of research at Huatai Securities in Hong Kong.
In most places, including Hong Kong or Singapore, a retail customer calling her bank must wait and listen through a long menu of options, often going through multiple key-ins before reaching a human being – whether for routine information or when facing an emergency. [While accompanied by some of the worst music in recorded history; seriously, who chooses this stuff? –Ed.]
But someone calling their bank in mainland China will almost immediately be greeted by a “customer service manager” who will answer your question. Increasingly, this responder is a robot powered by artificial intelligence.
“You can hardly tell if it’s a robot,” Chen told DigFin. “The answers given are relevant and helpful.”
As chatbots take on more responsibilities, for a growing slice of the market, banks are branding them with distinct personas.
For example, China Construction Bank has “Xiaowei”(小微), a chatbot serving its WeChat public account and mobile bank. Xiaowei is described as a 27 year-old Scorpio girl who is good at ballet and piano, and who likes to play yoga and badminton. Xiaowei’s favorite books are “Pride and Prejudice” and “Jane Eyre”. She is waiting to meet her true love.
Last year, CCB said Xiaowei had handled 1.9 billion customer interactions. The robot can understand more than 600 different versions of one question, and can handle more than 80 kinds of banking services.
It's easy to attract customers, difficult to keep them
CCB isn’t alone: most big Chinese banks have created lively images for their customer service bots, such as China Merchants Bank’s Xiaozhao (小招). Nor are banks alone: China Unicom has “Womei” (沃妹); travel company Ctrip has “Xiaoyou” (小游); and logistics company SF Express has “Fengxiaoman” (丰小满).
However, behind all of these characters, it’s the same robot, designed by Shanghai Xiao-i Robot Technology. This company has emerged as China’s leading A.I. customer service provider, with backing from investors like Alibaba, China Broadband Capital, China Everbright, and Huatai Securities.
Charming the customer
“Some of our early clients recognized that the talking style of these robots were similar to Xiao-i,” said Zhu Pinpin, CEO and co-founder of Xiao-i.
In 2004, Zhu and his partner, Yuan Hui, invented a chatbot for MSN and had attracted tens of millions of users. But they came to find the business model had bottlenecks to further growth.
China's automated banking customer service is far ahead
“It’s easy to attract customers but it’s difficult to keep them,” Zhu told DigFin. “After one or two weeks, the curiosity fades and clients go away if we don’t have added value...You can see the same problem for Apple’s chatbot, Siri. It needs music or other services to keep the clients.”
(The company has previously sued Apple in China for allegedly stealing some of the intellectual property behind its bots.)
Zhu’s team decided to add financial customer services on top of Xiao-i’s ability to converse.
As an early provider of A.I. to big companies, Zhu put in a lot of efforts to persuade banks and telecom companies to use it. Now, 80% of China’s 50 biggest banks have adopted Xiao-i’s solutions, for phone banking, online banking, or both.
The result is measurably better service for customers, and cost savings for the banks.
For example, according to Zhu, after Shanghai Pudong Development Bank adopted Xiao-i Robot, the communication time for each customer’s enquiry decreased from 127 seconds to 27 seconds. SPD Bank was able to save Rmb6 million per year for every 20-second reduction in the time to route a query to a human.
Human in the loop
Recently, iFLYTEK, a leading Chinese intelligent speech and language company, has been challenged for using simultaneous human interpreters to make its A.I. translator look good. The company calls its service a man-machine collaboration.
But it raises the question of how much value is coming from the software, and how much is still about humans manning call centers?
There is no A.I. that comes from nothing
Zhu Pinpin acknowledges there has been a lot of hype about A.I. that makes it out to be a lot more powerful than the reality. For example, translating voice from English into Mandarin remains difficult, especially in jargon-filled fields like finance. “Humans are still required to work with the machine,” he said.
This is called the “human in the loop”.
Zhu said, “There is no A.I. that comes from nothing.” He illustrates this with the example of China Merchants Bank, which introduced A.I. customer service in 2012.
It took a month to transfer enough knowledge to the machine and train it to answer various questions. After its accuracy reached 80%, CMB put it on a three-month trial before an official launch. But even now, the bank keeps the robot on a leash, with a team of dozens of people in the operations team tasked with maintenance.
But this headcount is nothing compared to the 3,000 or so that banks would need to provide the same degree of response without chatbots, Zhu says. But these are not call-center staff; they are back-office operations folk.
If the bot gets an unfamiliar question from a customer, it first tries to associate it with other questions in its experience, and asks a data analyst if it’s on the right track. Whether the answer is yes or no, the A.I. learns more about what’s relevant.
If the question is not something its database can immediately answer, the machine searches other websites and unstructured documents, querying its human overseers to see if it’s on the right track.
“A.I. needs external feedback and correcting to develop,” Zhu said. “Human judgment constantly passes new abilities to the machine.”
Hong Kong hurdles
So will other jurisdictions see a similar takeup of chatbots by banks? It’s early days, says Zhu, although Xiao-i has now signed up Bank of China Hong Kong.
Xiao-i Robot delisted from a stock exchange in China, the National Equities Exchange and Quotations Board, this past spring, due to insufficient revenues. It is reportedly seeking a listing in Hong Kong; Zhu would not comment on the company’s financial plans, but tells DigFin that the company should break even this year.
He likens Hong Kong’s current environment to the early days of Xiao-i Robot. Despite seeing chatbots go mainstream in mainland China, Hong Kong banks remain concerned about whether they have enough customers to justify the deployment, whether their I.T. systems are up to snuff, whether the Cantonese versions are adequate, and how local culture might receive Xiao-i-style bots.
Ryan Lam, head of research at Shanghai Commercial Bank (H.K.), says local banks would have to make extra investments to tweak mainland-derived software. Moreover, China’s banks are more profitable and have bigger capital bases they can allocate to such projects.
Banks will find that customer service driven by chatbots will become “hygiene”, that is, a vanilla, expected service, but not a differentiator. They face expenses to onboard Xiaoi-style chatbots, with little direct impact on attracting new depositors.
Tommy Ong, managing director at DBS in Hong Kong, says banks in Hong Kong, Singapore and elsewhere also have greater sensitivity to data privacy.
Xiaoi’s Zhu says banks can opt to train its machine without allowing the company see any of the data. But the less data they can harvest, the lesser the result. “Sharing data means better feedback, and faster adjustment,” he said. But he doesn’t think data privacy will remain a barrier to his selling his wares in Hong Kong.
“It’s common that banks are very cautious at the beginning,” he said.