When Ms Liu wanted to upgrade her home kitchen, the first tool she picked up was her smartphone.
Opening the app of Ping An Puhui, the consumer finance business, Ms Liu simply tapped in her basic personal information, answered a few questions from a representative from Ping An Puhui through the phone’s camera, and a loan of RMB60,000 was deposited into her account.
Ms Liu was delighted. The Ping An Puhui representative felt nothing, despite her importance: it is the first ever AI loans pre-assessment robot to be deployed on a large scale in the global credit market.
Ping An Puhui’s AI robot interacts with customers through online face-to-face interviews. With 500,000 standard Mandarin corpus and thousands of hours of training with humans, the robot is able to engage in natural-sounding conversations with humans and provide services comparable to that of a human representative.
For a loan pre-assessment, the robot will come up with interview questions based on personal data input by customers, including occupations and addresses. It will then collect and analyze the answers with other relevant data to pre-assess whether a customer qualifies for a loan.
Ping An has enabled loan assessment through video streaming since 2016 bringing the whole loan approval process online.
Now, the Ping An Puhui robot addresses the challenges of face-to-face assessments performed by humans, such as cost, subjective risk control judgement, inconsistent service quality and limited service hours.
During a loan pre-assessment interview, Ping An Puhui’s AI robot is doing a lot more behind the scenes while chatting. On top of automatic speech recognition (ASR), natural language processing (NLP) and dialogue management (DM), the robot is equipped with face recognition and micro-expression recognition technologies. It also uses anti-fraud technologies such as image recognition and multi-face/face-switching detection, which can flag if the applicant is with someone else in the room, potentially coaching them.
Ping An’s Puhui robot is groundbreaking, bringing together a range of innovative AI technologies. It records voice characteristics, lip shape and expressions of humans and mimics them for human-like interactions. As the robot interacts with more people, it learns and integrates more natural language.
At the same time, the system compares the image of the applicant with a database of faces of known high-risk borrowers. The robot, with the customer’s authorization, can also detect the surrounding background of customers to check for similarities in physical settings with previous cases of fraud.
The use of AI and big data does not replace the human involvement in the risk assessment process, but it is helping to increase the convenience and efficiency of the small loans process for customers, while enhancing the risk assessment process for the business.