Turn Every Resident Call and Email Into a Structured Case
A case engine that transcribes Cantonese calls and parses emails into structured, bilingual reports across a 150+ site portfolio.
- ThinkCol built a case engine that transcribes Cantonese calls, parses emails, and uses LLM extraction to produce a bilingual structured report in seconds.
- Each case is tagged against a bespoke 16-category Hong Kong property taxonomy and mapped to the exact site and block, with urgency, escalation risk, emotion, and tone flagged automatically.
- Across the portfolio, the system surfaces patterns no single case shows — issue contagion across blocks, silent sentiment escalation, and hidden repeat complainers across channels.
A major property management group fields thousands of resident and tenant calls and emails a month in Cantonese, across more than 150 malls, shopping centres, and residential estates. Each case was handled in isolation, with no way to see spreading issues, repeat complainers, or sites quietly sliding.
ThinkCol built a case engine that transcribes Cantonese calls, parses emails, and uses LLM extraction to produce a bilingual structured report in seconds: who and where, urgency and escalation risk, emotion and tone, contact type, and a bespoke 16-category Hong Kong property taxonomy mapped to the exact site and block.
Across the portfolio, the system surfaces patterns no single case shows on its own — issue contagion across blocks, silent sentiment escalation, and hidden repeat complainers across channels — with plain-English querying on top. It runs in the client's own secure Hong Kong cloud and was independently graded for accuracy before go-live.
Why customise? Off-the-shelf tools are English-first, log tickets without connecting them, and lack a Hong Kong property vocabulary.
