Theron Li

A leading Hong Kong specialist clinic group needed to modernise operations during rapid expansion. The engagement grew from document processing into an AI-enabled CMS rebuild.

• Mapped medical, financial and operational document workflows and selected recognition and validation as the initial AI use case. Led an on-premises AI-OCR rollout, integrating recognition, validation and filing into daily workflows.
• Built on the initial engagement to drive a clinic management system (CMS) rebuild, defining scope, product architecture and delivery phases. Led the team in using AI-assisted coding to map the existing SaaS system’s business logic, upgrade the architecture and integrate AI into key workflows. The team completed core functionality across six operational areas and 20+ modules in 2–3 months, including reception, consultations, pharmacy, laboratory and billing, and conducted cross-role workflow validation.

Healthcare and trade-document projects had different requirements, making one-off customisation costly. The product needed a reusable core that could accommodate industry-specific needs.

• Mapped shared and industry-specific requirements across projects, turning document recognition, data linking, validation and workflows into reusable components, with configuration and client-specific integrations handling the differences.
• Developed industry solutions from delivery experience, clarifying what could be reused and what required new development in subsequent engagements, and supported expansion into further industry use cases.

Owned solution design and demo development for a tax engagement where AI use cases and implementation scope were initially unclear.

• Combined client discovery, tax-process research and industry best-practice analysis to identify pain points and priority use cases, then designed the solution architecture and integration approach.
• Built a demo of the proposed workflows and used scenario walkthroughs and client feedback to translate broad expectations into concrete processes, progressively clarifying requirements and implementation scope.

Mobility support covered both information requests and service actions, requiring clear routes between automated and human support.

• Built intent-specific agent workflows with MCP tools for determining responsibility, issuing coupons, querying orders and accessing user tags. Used real question–answer pairs to create test cases and benchmarks, identifying failures in knowledge, responses and workflow logic to guide improvements.
• Designed support entry points, in-app service cards and escalation paths as a single customer journey. Defined responsibilities and handoff criteria for AI, support agents following standard operating procedures, and specialist agents, routing each request to the appropriate support tier.

Managers varied in analytical experience and approach. They needed a consistent way to spot performance issues, investigate causes and decide what action to take.

• Interviewed managers at headquarters and regional offices, mapping three business domains and 14 core KPIs and their drivers into a common framework for performance tracking and root-cause analysis.
• Designed and launched an executive dashboard for tracking targets, diagnosing exceptions and recommending actions, turning the framework into a day-to-day management tool. After launch, over 95% of target managers actively used the executive dashboard.

• Drew on frontline research and claims analysis to define drink-driving detection needs, assess relevant in-house technology and identify risk indicators and detection rules. Developed the product solution and coordinated with technical teams to launch the model and refine it over time. The model contributed over RMB 10 million in annual loss reduction.