JapanFinance and insurance2024 – ongoing

Mitsubishi UFJ Trust and Banking

Generative AI answers to internal questions on financial market transactions

Mitsubishi UFJ Trust and Banking cut specialist response time by 50% with a generative AI that answers internal market-operations questions

WorkingCompany's own claim

Problem

Staff across the bank sent questions about financial market transaction operations to specialist departments. Answering them took expert knowledge and department time.

Approach

The bank ran a proof of concept with the AI company Kasanare and put the system into business use in April 2024. The generative AI reads system specifications, operation manuals and FAQs, and it partly automates the answers to these internal inquiries. The bank plans to reuse the prepared data for future AI agents.

Results

over 90%answer accuracy in the proof of concept, as reported by IT Leaders from the bank's announcementCompany's own claim[1]

50%reduction in response time for the department handling the inquiries, per the bankCompany's own claim[1]

several hundred thousand hours or more per yearexpected annual saving if the system is extended to other areas, the bank's estimate and not a resultCompany's own claim[1]

Takeaway

The bank started with one narrow, knowledge-heavy inquiry flow and built the document base that later projects can reuse.

Sources

三菱UFJ信託銀行、金融市場取引に関する社内問い合わせへの対応に生成AIを適用
IT Leaders5 Aug 2024News article

Source summaries

[1]IT Leaders5 Aug 2024

IT Leaders reported the bank's announcement of 2 August 2024. The system, built with Kasanare, reads specifications, manuals and FAQs to answer internal questions on market transactions and has been in operation since April 2024. The article gives over 90% accuracy in the proof of concept, a 50% cut in response time and an estimate of several hundred thousand hours saved per year if extended.

Full details are in the original sources.

Added 2 Nov 2026 · Updated 2 Nov 2026