Newsroom  •  December 01, 2019

Start the Conversation

Honeypot Field to Catch Bots
Honeypot Field to Catch Bots

Expanding Use of AI in Bank Operations: AI Document Information Extraction & Chatbot Lets Bank Staff Handle Specialized Tasks

A growing number of banks in Japan are moving to improve the efficiency of their operations by using artificial intelligence (AI). Recently, Mitsubishi UFJ Trust and Banking Corporation (MUTB) developed an AI solution that can import an enormous number of documents, including contracts, class action complaints and mutual fund prospectuses, and automatically extract only the information needed by clients. The bank will implement a system using this technology in 2019 that will allow it to reduce the time spent on a variety of internal operations by 5,000 hours annually. With this and other solutions, such as a chatbot system that automatically responds to work-related questions, banks can have their employees engage in more specialized tasks.

In the past year, MUTB, a subsidiary of the Mitsubishi UFJ Financial Group (MUFG, 8306), introduced an AI document reader developed by a US startup. While originally meant for lawyers, the AI was reworked for financial use and support added for the Japanese language. By repeatedly training the AI to extract indicated portions of a document, it eventually is able to obtain the same sections from other documents on its own. After training on about 30 samples, the AI is able to pull out information with a high level of accuracy and drop it into a spreadsheet.

MUTB introduced the solution at The Master Trust Bank of Japan, its asset management subsidiary. It was used when inquiring whether client investors wanted to associate themselves with a shareholder class action brought against a specific company they hold stock in. The AI goes through the class action complaints, which are hundreds of pages long, and pulls out the relevant holding periods for shares, share types and filing deadlines. Something that took humans eight hours to finish, the AI could do in four. However, as people tended to do this alongside other tasks the entire process actually took around three days. As it receives an average of 14 complaints per month, MUTB figures that this will translate to 672 fewer hours spent annually going over complaints.

Use of AI at MUTB will be expanded to around 10 different operations, including asset audits performed when forming new funds and FX risk identification. This will lead to a workload reduction of 5,000 hours annually in 2019. Based on the results, there are plans to expand the introduction of AI across the MUFG Group.

Sumitomo Mitsui Banking Corporation (SMBC), a member of the Sumitomo Mitsui Financial Group (8316), has introduced an AI chatbot that answers employees’ queries. While originally meant to handle questions about IT systems and HR procedures, the chatbot was updated so it could also give answers about FX derivatives and other financial products starting from November 2018. An English version went online in June 2019.

The AI selects responses appropriate to the current situation having analyzed a variety of data including a collection of past answers made by humans. It was fine-tuned to provide correct responses even if questioners use different inflections or phrasing.

The Mizuho Financial Group (8411) has plans to offer a system to regional banks that can automatically read handwritten wire transfer forms. It makes decisions by comparing what is written on the forms with account information registered at the banks. Using AI to distinguish between the different formats used by companies increases form processing speed by a factor of 10. This system will be introduced at the Group’s Mizuho Bank within this fiscal year.

Banks across Japan are moving ahead with efforts to cut costs by reducing workloads. MUFG Bank has set a goal for itself to phase out more than 10,000 jobs by 2023. While it is introducing robotic process automation (RPA) to automate routine tasks, utilizing AI means that it can entrust more advanced work to RPA. As the number of visitors to branches continues to decline with the spread of Internet banking, bank employees have less office work to do. Due to this, banks will be assigning their human resources to handle highly specialized tasks such as providing advice related to individual asset management. (Shiko UEDA)

The Number of Areas Where AI Is Used Is Growing:

  • Mitsubishi UFJ Trust and Banking - OCR reading and extraction of information from large amounts of documents
  • SMBC - Answering employees’ questions through chatbot system
  • Mizuho Financial Group - OCR reading of wire transfer forms
  • Mizuho Bank - Small and medium business loan screening
  • MUFG Bank - Small and medium business loan screening
  • Mizuho Bank & SoftBank - Personal loan screening
  • Customer services
  • Internal banking operations