← Cases
JapanFinance and insurance2025 – ongoing

Hyakugo Bank

Generative AI and AI agents for housing-loan pre-screening documents

Hyakugo Bank cut housing-loan document handling from about 20 minutes to 7 minutes or less in a Hitachi trial

In progressVendor case studyVendor ↗

Problem

Housing-loan pre-screening depended on staff reading non-standard documents such as property materials and floor plans. They typed the details into the bank's system and checked them by hand.

Approach

Hyakugo Bank and Hitachi tested generative AI that reads the documents and enters the data into the bank's internal system. AI agents then compare the entered data with the application and flag inconsistencies and risks for loan officers, who make the final decision. The trial ran from August 2025 to February 2026. The bank plans to apply the system to housing-loan pre-screening from fiscal 2026.

Results

7分以下time per case at or below 7 minutes in the trial, down from about 20 minutes, per HitachiVendor case study[1]

Takeaway

The result comes from a trial, and the bank had not yet started live use when the release came out. Loan officers keep the final decision.

Categories

Business function
Production and quality
Innovation type
Process innovation
Evidence
Vendor case study

Sources

Source summaries

[1]Hitachi27 Feb 2026

Hitachi's release describes a trial with Hyakugo Bank from August 2025 to February 2026. Generative AI read non-standard housing-loan documents and entered data, and AI agents checked consistency. Time per case fell from about 20 minutes to 7 minutes or less. The bank plans to start applying it in fiscal 2026.

Full details are in the original sources.

Added 2 Nov 2026 · Updated 2 Nov 2026
Open case page ↗