Annex Chatbot
A multi-functional AI conversational platform supporting multimodal interaction and custom script extensions, delivering seamless automated communication experiences for enterprises.
https://annex-chatbot.2dqy.com/
Visit BobAI
Automation
Chatbot
Customer Support
NLP
Project Overview
I. Core Use Case: Embedding as a Government Subsidy Consultation Endpoint in Hong Kong
This section demonstrates how to embed AI-powered recognition capabilities into the Hong Kong SAR Government’s subsidy service ecosystem, transforming complex policy documents (e.g., Application Guidelines, Compliance Handbooks) into interactive, real-time data flows.
| Use Case Category | Core Logic & Technical Implementation |
|---|---|
| Intelligent Policy Intent Routing | Delivers chat-automation capability for platforms like Subsidy Connect. Automatically detects user identity (e.g., local enterprise holding a Business Registration Number) and intent, then precisely routes them to specific subsidy pathways — such as BUD Fund, TVP, or EMF. |
| Eligibility Criteria ETL Pipeline | Builds an end-to-end chain: User Q&A → Eligibility Factor Extraction → Structured JSON Fields. Automatically identifies key data points (e.g., annual turnover, MPF contribution records) to pre-populate official application forms from HKPC or relevant policy bureaux. |
| Dynamic Knowledge Base Interaction | Leverages the endpoint to fetch the latest subsidy terms (e.g., 2026 updates to e-Commerce Easy), significantly improving AI accuracy in automated eligibility checks, ensuring recommended schemes align with current policy frameworks. |
| Cantonese Semantic Mediation Layer | Processes colloquial queries (e.g., “Can I get funding to start an online store?”, “I want to exhibit in mainland China”) through secondary linguistic normalization, mapping them to standardized policy terminology (e.g., TVP IT Funding Scheme, BUD Special Fund) and outputting clean, structured data. |
II. Business Value: Real-World Deployment for Local Subsidy Service Platforms
This section focuses on resolving pain points faced by Hong Kong enterprises during subsidy applications — directly empowering execution agencies like the Hong Kong Productivity Council (HKPC), trade associations, and professional advisory teams.
1. Automated & Scalable Consultation Efficiency
- Guided, Decentralized Self-Service: Uses branching logic (Q1–Q3) to guide users through preliminary eligibility screening — drastically reducing operational overhead for frontline staff handling repetitive initial queries like “Is my company eligible?”.
- Multilingual Subsidy Navigation: A single logic layer supports seamless switching among Traditional Chinese, Simplified Chinese, and English — enabling both local SMEs and foreign-invested firms in Hong Kong to quickly identify and access schemes such as PAG or CMDF, boosting conversion rates.
2. Professional Terminology Alignment & Compliance Governance
- Rule-First Matching Engine: Enforces predefined policy rules strictly (e.g., non-local salaried employees cannot be counted as R&D personnel under TVP). The AI auto-generates precise contextual descriptions, guaranteeing 100% accuracy in policy interpretation, eliminating misguidance to applicants.
- Closed-Loop Review Pathway: Implements a “Conversation-as-Initial-Review” mechanism. Administrators can instantly audit Phase 1/2 dialogue JSON logs, rapidly assessing applicant authenticity and compliance risk — reducing downstream audit burden and exposure.
3. Data Collaboration & Secondary Conversion
- Automated Lead Capture: Offers one-click export of dialogue JSON, seamlessly feeding qualified leads — those who have completed preliminary eligibility checks — into CRM systems or advisory teams, shortening the business cycle from consultation to formal application submission.
- Real-Time Event Notifications (Webhook): Triggers instant alerts to advisors when users express high-match intent (e.g., alignment with the 2026 New Industrialisation Subsidy), enabling timely human intervention at the critical ‘golden window’ of subsidy application.