Public Administration AI (Qurify-Based)
✓ In production · Hongcheon County since May 2026AI Civil Complaint Management System
Qurify ontology-based · closed-network LLM support across the entire complaint handling process
Complaint analytics, AI-powered staff assignment, automated answer drafts, and automatic retrieval of similar cases and statutes — all running inside the government's closed network with no external connection.
At a glance
One-sentence definition
The AI Civil Complaint Management System is a closed-network, on-premise Qurify product for public complaint work — intake, assignment, draft answers, similar cases, and statute lookup.
Who it is for
Municipal complaint desks, handling departments, audit and statistics officers.
Problem it solves
Repeat complaints are classified by hand, similar cases and statutes take time to find, and answer quality drops when staff change.
Key capabilities
- ·Complaint analytics and AI staff assignment
- ·Automated answer drafts
- ·Similar-case and statute lookup
- ·Role-based screens and operations statistics
Environment, integrations, deployment, security
Closed-network government LLM, on-premise, no external internet. In production in Hongcheon County since May 2026.
Limitations
Limited to public-administration complaints. It is not a private contact-center product. Metrics are from Hongcheon operations and vary by institution.
Verified Performance
Verified Performance Metrics
5
min
AI answer drafts pre-generated overnight (per case)
5
sec
Response time for similar-case and statute search
650
users
Concurrent users verified (spec supports up to 800)
8/8
features
All 8 core features passed testing and final inspection
Business Impact
Impact — Before vs After
| Area | Before | After |
|---|---|---|
| Assignment | Opening complaints one by one to assign officers manually | AI recommendation → assigned with a single click |
| Answer drafting | Manually searching past cases and statutes, then writing (tens of minutes) | Review AI draft → revise via chat → consult the similar-case and statute panels |
| Legal grounds | Separate searches on the national law portal | Relevant statutes and provisions surfaced automatically in a dedicated panel |
| Similar cases | Keyword search, relying on memory | Summaries and evidence from past cases surfaced automatically in a dedicated panel |
| Deadline tracking | Checking due dates individually by hand | Automatic D-day badges with color-coded warnings |
Reduced to below 80%
Manual work rate
3–4 days → within 1 hour
Complaint response time
~40x increase
Cases handled per officer
95% (processed within 4 days)
Deadline compliance
* Based on the Hongcheon County deployment
Case Study — Hongcheon County
- Client
- Civil Affairs Division, Hongcheon County, Gangwon State
- Project period
- Oct 24, 2025 – Apr 30, 2026 (189 days)
- Users
- ~650 public officials
- Complaint volume
- ~10,000 complaints per year
- Data construction
- 8,000 complaints (2016–2025) + statutes and ordinances + 746-person org chart
- Integrations
- End-to-end Saeol SSO integration, daily batch sync with e-People
- Verification
- All 8 core features passed pilot operation and final inspection → full production since May 2026
- Maintenance
- Monthly on-site visits, AI model recalibration every 6 months, quarterly statute updates
- Expansion
- Ready to scale as a standard for local governments nationwide — one closed-network unit plus an optional internet-network deployment
“Together with Hongcheon County, we successfully built an AI system that runs reliably even in a fully closed government network.”
“Speed AI Min Hongcheon goes beyond faster complaint processing — it helps ensure residents receive stable, consistent service.”
Core Features
8 Core AI Features
AI supports the entire process — from intake and assignment to answer drafting and completion.
AI Staff Assignment Recommendations
Automated nightly analysis plus real-time recommendations. Presents the top 3 recommended officers along with the reasoning behind each assignment
AI Answer Drafts + Chat Revision
Drafts pre-generated overnight (within 5 minutes per case) — ready when staff arrive. Request revisions via chat for instant regeneration, with 4 revision types, undo, and copy
Similar Complaint Search
AI similarity search automatically surfaces past complaints with 50%+ match. View full text of previous answers and handling departments
Automatic Statute & Ordinance Lookup
Automatically matches national statutes and local ordinances. Displayed with separate badges; click to view the full provision text
Real-Time Monitoring Dashboard
Alert cards for same-day and approaching deadlines, sorted by due date. Processing status updates in real time
Completion Workflow + 4-Role Permissions
Two resolution paths: complete or reject. Four permission tiers — assigner, handler, administrator, and system owner
Unified Handling of Offline Complaints
Complaints received by fax, phone, or walk-in get the same AI-powered processing. Unusual complaints are flagged automatically
Statistics & Reports
Processing statistics by officer, department, and period, with Excel export. Visualizes reductions in processing time
System & Security
System Architecture & Security
An on-premise architecture optimized for closed government networks guarantees both security and performance.
Deployment
On-premise installation inside the government's closed network. All processing completes without any external internet connection.
Data Security
Complaint data can never leave the network. Security measures documentation and network diagrams were submitted in the Ministry of the Interior and Safety's official format and approved.
AI Engine
Proprietary lightweight LLM deployed on-premise (GPU workstation). No external LLM APIs, no token costs.
Verified Integrations
Saeol administration system SSO auto-login, daily batch sync with the national e-People petition system
Extensible Integrations
Government24, SEUMTER, Bokjiro, next-generation vehicle management, geospatial systems, and more (when the operating agency provides APIs)
Frequently Asked Questions
Questions actually asked by other local governments.
Q.We can't use external cloud services. Is this still possible?
Yes. The system is installed and operated directly on servers inside your closed government network. Our proprietary LLM runs on on-premise GPUs, so no external LLM APIs or internet connection is needed at all. Complaint data can never leave the network.
Q.Do we need to modify our existing complaint systems (Saeol, Eungdapso, etc.)?
No. The AI functions are developed separately and connected via API, without touching your existing systems. Saeol SSO auto-login and e-People integration are already verified in production. After the benefits are proven, features can be gradually merged into your existing system's UI.
Q.Besides e-People, what systems can it integrate with?
e-People (batch) and Saeol (SSO) integrations are already built and verified. Government24, SEUMTER, Bokjiro, the next-generation vehicle management system, geospatial systems, and others can be integrated when the operating agency provides APIs. Each system involves its own administrative procedures (official requests, etc.).
Q.Are AI answers generated in real time or in batches?
Hongcheon County uses overnight batch generation, so drafts are ready when staff arrive in the morning. New deployments can be designed for real-time generation from the start. Similar-case and statute searches respond within 5 seconds on screen.
Q.Can it handle our agency's complaint volume?
Yes — from small to very large environments. We size the GPU infrastructure to your scale, and large deployments scale out with multiple GPUs in parallel. We provide an exact specification after reviewing your daily complaint volume, media types, and concurrent user counts.
Q.What about complaints with image or video attachments?
We recommend starting with text complaints, then expanding to images (OCR/image recognition) and video in stages. Image processing requires only a higher base GPU specification; video increases GPU costs significantly, so we review it separately based on volume and frequency.
Q.How is this different from a generic chatbot or LLM solution?
Rather than simple LLM answer suggestions, the system works on top of an ontology (Qurify) that precisely models your agency's workflows and data. Duty assignments, org charts, and statutory frameworks are structured as metadata, enabling accurate assignment, answers, and evidence. And it runs in a closed network with no external connection.
Q.What if our historical data is sparse or messy?
Data cleansing and ontology construction are core phases of the project. Even with limited data, we analyze and refine what you have and convert it for AI training. As resolved cases accumulate in production, retraining continuously optimizes the AI for your agency.
Q.How is personal information protected?
Access is restricted by role (four tiers: assigner, handler, administrator, system owner), and personal information is masked during preprocessing before data is loaded. Masking scope and policy are designed in advance to match your agency's privacy standards. We have a track record of submitting and passing security documentation in the Ministry of the Interior and Safety's official format.
Q.What are the deployment process and timeline?
The standard deployment takes about 6 months: initiation and analysis (weeks 1–4), data construction (weeks 5–10), system development (weeks 7–16), installation and testing (weeks 17–20), pilot operation (weeks 21–22), and full operation (weeks 23–24). For competitive tenders, we can help draft the functional specifications for your RFP.
Timeline
Deployment Process
The standard deployment takes about 6 months (24 weeks), proceeding in stages. For competitive tenders, we can help draft the functional specifications for your RFP.
Initiation & Analysis
Weeks 1–4
Data Construction
Weeks 5–10
System Development
Weeks 7–16
Installation & Testing
Weeks 17–20
Pilot Operation
Weeks 21–22
Full Operation
Weeks 23–24
AI Civil Complaint Management System Brochure
Get the official brochure covering system features, the closed-network architecture, and the Hongcheon County case with verified metrics.
Considering the AI Civil Complaint Management System?
We'll design a configuration that fits your agency's environment, guiding you step by step from demos to PoC to production.