JapanFinance and insurance2025 – ongoing

Norinchukin Bank

Super RAG pilot for internal regulations and manuals

Norinchukin Bank reported 80–90% answer accuracy in a Super RAG pilot

In progressVendor case studyVendor ↗

Problem

Employees needed information spread across multiple regulations and manuals. Earlier retrieval-augmented generation testing produced answers too inaccurate for the bank's needs.

Approach

The bank tested Cinnamon AI's Super RAG on internal questions, starting with HR inquiries such as parental leave. It compared performance with earlier testing on the same use case. Wider deployment remained under consideration.

Results

80–90%pilot answer accuracy, compared with about 30% in earlier testing on the same use case, according to Cinnamon AI's customer interviewVendor case study[1]

Takeaway

The accuracy comparison came from a proof of concept. Reduced inquiry volumes and staff time remained expected benefits.

Sources

農林中央金庫が選んだ「Super RAG」——生成AI活用の最前線
Cinnamon AI22 May 2025Vendor publication

Source summaries

[1]Cinnamon AI22 May 2025

An April 2025 customer interview reported pilot accuracy of 80–90%, versus about 30% in earlier testing. The bank described HR inquiries as an initial use case and broader applications as future work.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026