evidoria

← Back to browse

Good practice Imported

Busan's AI Agent Service — A Four-Month Pilot to Triage Civil Complaints in Dialect, With No Results Yet Published

South Korea · Busan · See the South Korea profile · See the Busan profile

Top 100% 13/100 · Ask Evidence Copilot about this practice

Busan ran a 2025 pilot AI Agent for its "120" complaints hotline (219,108 calls in 2024; 108,703 in H1 2025), using a Busan-specific LLM with RAG to classify, summarize and translate dialect. Ten firms competed for the contract; no results were published by the pilot's end.

Busan's AI Agent Service — A Four-Month Pilot to Triage Civil Complaints in Dialect, With No Results Yet Published

Details

Promoter
Busan Metropolitan City / VAIV Company
Period
September–December 2025 (pilot)
Keywords
citizen services, generative AI, civil complaints handling

Description

In September 2025, Busan Metropolitan City, in partnership with VAIV Company, launched a four-month pilot of an "AI Agent Service" to handle its "Busan Civil Complaint 120" hotline, which took over 200,000 calls in 2024 (219,108) and 108,703 in the first half of 2025 alone — most of them routine questions from elderly residents about subsidy timing, eligibility and application steps. The project was selected through South Korea's Ministry of Science and ICT public contest for "Support for the Use of Hyper-Scale AI-based Platforms," with 10 companies competing for the development contract before VAIV was chosen.
The system pairs a Busan-specialized large language model with retrieval-augmented generation to classify incoming complaints, summarize their content, convert Gyeongsang-do dialect into standard Korean, and draft type-specific responses for staff to review. City officials identified dialect comprehension as the decisive risk factor: before committing to their own build, they examined neighbouring Gwangju's "AI Night Duty Officer" system (running since August 2024) and judged it insufficiently robust for Busan's dialect, choosing to develop a more tailored model instead.
Both the city's own announcement and local newspaper coverage of the pilot are explicit that its goals — easing staff workload, improving service quality and speeding up processing — are expected benefits, not measured results: no data on complaints actually handled by the AI, staff time saved, or citizen satisfaction had been published as of the pilot's scheduled December 2025 end date, illustrating a pattern common to 2026 government AI pilots, where OECD survey data show most agencies have started an AI project but only a small fraction reach measured, stable production.

Read the full analysis: https://www.busan.go.kr/eng/ai-translated-press-releases/1696100

Implementation

Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful