← Cases
South KoreaFinance and insurance2026 – ongoing

Kiwoom Securities

AI moderation of investor-community messages

Kiwoom Securities paired AI chat classification with human review as its supplier reported a system error rate below 5%

WorkingVendor case study

Problem

Kiwoom's live investor community generated more messages than staff could review manually. Moderation also needed to follow internal rules and financial-service security requirements.

Approach

Dalpa supplied AI monitoring for the community introduced in the Yeongungmun S# mobile application in January 2026. The system classified messages against Kiwoom's operating guidelines. Dedicated operators performed a second review after AI classification.

Results

below 5%overall system error rate reported by supplier Dalpa; classification accuracy was not separately disclosedVendor case study[1]

Takeaway

Human review remained part of moderation. The reported system error rate was not a disclosed measure of classification accuracy.

Categories

Business function
Customer service
Innovation type
Process innovation
Evidence
Vendor case study

Sources

Source summaries

[1]ZDNet Korea21 Jan 2026

Dalpa described supplying live AI message classification followed by operator review. It reported an overall system error rate below 5% during operation, including periods of concentrated traffic.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026
Open case page ↗