From application platforms used by people across the country, to data hubs that span systems, to LINE voice visit records in rural communities, we work with central and local agencies to turn policy into services that work.
The system serves the public nationwide, case officers in 22 cities and counties, partner financial institutions and the central authority. Records are linked all the way back to 2007, preventing duplicate subsidies and sparing agencies the back-and-forth of official letters: infrastructure at national scale. Subsidized households grew from about 150,000 in 2020 to more than 800,000 in 2026.
Data from household registration, land registry, tax and social subsidy agencies is consolidated under each applicant's identity. Case officers see a person's applications, supplementary documents, disbursements and recoveries over the years on one screen, without querying each source separately.
AI first identifies attachment types, flags what's missing and drafts plain-language requests for supplementary documents; the public is notified only after a case officer confirms. Subsidy amounts and eligibility are always decided by rules, never by AI.
Hello, your rent subsidy application is missing a copy of your household register. Please upload it within 14 days (by Nov 25), or bring it to your district office. We’ll continue the review once it’s in.
Decision logic has been pulled out of the code into five layers of parameters: regulation, policy, city/county, workflow and presentation. Every change carries an effective date and version history, and applies to all three subsidy types at once.
| Area | Outcome |
|---|---|
| Scale | Handles 1 million cumulative applications; read/write separation keeps peak-time queries from slowing intake |
| Review efficiency | Attachments are pre-flagged and supplementary-document requests drafted automatically, so notices go out as soon as a case is opened instead of after a full read-through |
| Management decisions | Cross-criteria statistics come from a single query instead of manually reconciling several exported reports |
| Policy agility | Policy changes are made through parameter settings instead of code, and cities and counties can maintain their own local rules |
Most community volunteers in Chaozhou, Pingtung are seniors themselves, and visit notes used to live on paper and in memory. Now they record the conversation in LINE, AI turns it into a record of physical condition, emotional state and follow-up, and nothing is saved until the volunteer confirms. Volunteers spend their time with elders, and case officers spot who needs help sooner.
Volunteers used to write up visits from memory afterward. It took time, the detail varied, and changes in an elder’s health or needs were easily missed. Now the conversation is recorded during the visit, AI produces a transcript and organizes it into physical condition, emotional state and follow-up items.
AI output can contain mistakes, and saving it unchecked could affect later care decisions. So AI results always start as a draft, and an official record is created only after the volunteer confirms. Case officers can review and correct it again in the back office.
Visit records touch on elders’ health and family situations, and volunteers work on their own phones. The platform separates data access by role and masks elders’ names on volunteers’ phones.
Over three years, the Maritime and Port Bureau is taking the training, testing and certification of Taiwan’s maritime personnel fully digital. As the program’s project management and technical advisory team, we keep every sub-project’s schedule and specs aligned, hold a high bar for technology and security, and work alongside training institutions so the new tools become part of everyday teaching.
Read the case→
Connects dozens of previously siloed systems, including medical administration, dedicated pediatric physicians and hospital accreditation; cleanses and standardizes the data to the FHIR international healthcare standard; and links external data on population, national health insurance and long-term care. Sensitive data is de-identified first and processed in the agency's dedicated data center. Accepted and closed for 2024, renewed for 2026.
Read the case→Connects roughly 80 to 100 data sources into a regularly updated industry database. A built-in AI Agent drafts briefing notes with cited data sources in minutes, and interactive charts support filtering and drill-down on tablets and phones.
Read the case→Cleanses hundreds of millions of IoT sensor records and turns them into information the public can understand. Working with the TAIWAN AI RAP platform, it connects TAIDE and open-source models while the data stays in the agency's data center and is reached only through an API.
Adopting AI doesn't change who is accountable. Every project follows these four rules.
Answers draw only on laws and agency documents; if the basis is insufficient, no answer is given.
Any judgment on subsidy amounts or eligibility is handled by rules, never by AI.
Can be deployed in the agency's own environment; data is never used for model training.
Query conditions, AI outputs and human edits are all logged and traceable record by record.
| Agency | Period | Scope |
|---|---|---|
| Maritime and Port Bureau, MOTC | Since 2026 (ongoing) | Smart Digital Upgrade Program for Maritime Personnel: project management (PMO) and technical advisory, cybersecurity oversight, digital transformation support for seafarer training institutions |
| National Land Management Agency, Ministry of the Interior | Since 2026 (ongoing) | Housing Subsidy Review and Online Application System: feature expansion, AI adoption, on-site customer support and nationwide training |
| Department of Medical Affairs, Ministry of Health and Welfare | Since 2024 | Medical Administration Digital Transformation Project (accepted and closed for 2024, renewed for 2026); Medical Workforce Management Efficiency and Digital Application Project |
| Ministry of Economic Affairs | Since 2023 | CRM platform build and operation, policy information outreach, registration and contact list analytics |
| Small and Medium Enterprise and Startup Administration, MOEA | Since 2026 | Business service outreach, "Ma Shang Ban" CRM, data integration and performance analysis |
| Ministry of Environment | Since 2024 | CRM platform, external API and data sync, environmental data API integration |
| National Center for High-performance Computing (NCHC) | Since 2025 | Public IoT AI data dashboard, knowledge management platform |
| Taoyuan City Government Economic Development Bureau | Since 2024 | Industry data analytics platform |
| Kaohsiung City Government Economic Development Bureau | 2023–2024 | Contact list management system, event check-in scan database |




