The neighborhood clinic near your home: the doctor completed a diabetes risk assessment in 25 seconds that used to take 20 minutes. The FamilyMart clerk tapped for three minutes to handle what used to take fifteen minutes of fresh-food ordering. The customer service that responded to your LINE message wasn't a person. Taiwanese people interact with AI more than a dozen times a day — and almost no one realizes it.
A Doctor's 25 Seconds
Dr. Chen is a family medicine physician at a community clinic in New Taipei City, seeing forty to fifty patients a day. Starting in March 2026, his consultation workflow gained a new step: as he opens a patient's record, the National Health Insurance Administration's AI diabetes risk assessment system automatically runs a complications risk classification. A color block appears on the right side of the screen — green for low risk, yellow for medium, red for high. Twenty-five seconds.
Three months earlier, the same assessment required manually reviewing medical history, cross-referencing lab data, and consulting clinical guidelines — at least 20 minutes. Even earlier, most primary care clinics didn't conduct this assessment at all, because there was no time.
The system is called "AI-on-DM," and it is the fruit of a five-year collaboration between the NHIA and Google Health, deployed to approximately 20,000 primary care clinics across Taiwan. What originally required 40 specialists spending three weeks to screen 20,000 people can now be completed in 1 hour and 24 minutes — a 14,400-fold increase in capacity. That same month, the NHIA launched a Gemini-powered health education assistant inside the "Health Bank" (健康存摺) App used by tens of millions — Taiwan may be the first country globally to integrate a large language model into a national public health app.
Dr. Chen says his biggest feeling is not "it's faster," but "things that were previously impossible are now possible." A community clinic cannot afford to maintain a specialist team for risk assessment, but AI gives every clinic that capability.
The AI Inside the Convenience Store
At 6:30 in the morning, a FamilyMart store manager opens a tablet. The screen already shows today's recommended fresh-food order quantities. Weather, foot traffic, yesterday's sales, nearby events — the system has factored all of it in. Adjust two items, press confirm: three minutes. Three years ago this took fifteen minutes, guided by experience and intuition.
FamilyMart implemented an AI fresh-food ordering system in 2022. Ordering time was reduced by 75%, and fresh-food waste dropped by close to 10%. Combined with the "Friendly Eating Hours" 30% discount mechanism for items approaching expiry, the system reduces approximately 370 metric tons of food waste per month. President Chain Store (7-Eleven) has similarly implemented a sales-forecast ordering system, taking inventory management across more than 7,000 nationwide stores from "going by feel" to "looking at data."
Taiwan's convenience store density is the second highest in the world — one store per 1,500 people. When AI permeates this distribution channel, it reaches not early technology adopters but everyone who walks into a store to buy an onigiri.
LINE: Taiwan's Invisible AI Gateway
Taiwan's AI penetration didn't arrive from Silicon Valley — it grew from inside LINE group chats.
LINE's monthly active users in Taiwan reach 22 million, with a penetration rate exceeding 90% — higher than Japan or Thailand. Medium-to-heavy users open LINE an average of 22 times per day. This is not just a messaging app; it is the primary interface through which most Taiwanese people access digital services.
By late 2024, the number of official LINE accounts in Taiwan exceeded 3.09 million, with one-to-one conversation messages increasing 45% year-on-year, and notification-type message volumes up 4.5 times year-on-year. The driving force behind this is AI: LINE official accounts now have built-in AI chatbot functionality — businesses just need to upload product data or URLs and the system can automatically generate FAQs and respond to customer questions in real time. Chatbots that previously required engineers to integrate can now be activated from a back-end dashboard in a few clicks.
In October 2025, LINE Taiwan announced at its CONVERGE annual conference entry into the "AI Agent era," formally launching MINI HOME as a new user gateway, integrating multiple brands' LINE MINI Apps within the LINE Wallet page. In the first half of 2026, LINE Touch is also set to launch — holding a phone near an NFC tag to instantly activate a service page, with coupons, ordering, and event sign-ups all completing in a single tap.
Translation bots are also everyday tools in the LINE ecosystem. Third-party translation bots such as Ligo support more than 100 languages and are widely used in LINE group chats in tourism and international trade. And LINE's CLOVA voice recognition technology has been localized for Taiwan in Chinese, able to convert speech to text records even in noisy environments — applied to meeting minutes, customer service voice analysis, and similar scenarios.
Taiwanese people interact with AI more than a dozen times a day — it's just that the AI is hidden behind LINE's interface, and looks like "automatic system replies."
Another Side of Medical AI: From Colonoscopy to Bone Marrow Smears
The NHIA's diabetes model is only the tip of the iceberg. The "real-time AI polyp detection for colonoscopy" system developed by aetherAI in collaboration with National Taiwan University Hospital and Cathay General Hospital has received medical device approval from Taiwan's Food and Drug Administration (TFDA), with an accuracy rate of 95.8% and latency below 0.2 seconds. The training data comes from more than 3,000 patients and over 400,000 images. Their AI classification and count system for bone marrow smears has simultaneously obtained both TFDA and European CE certification — the first such case globally.
Taiwan's universal National Health Insurance system has accumulated more than 20 years of complete medical records covering 99.9% of the population. This data goldmine is being re-mined by AI. The model is based on the complication severity index certified by three major professional associations — including the Diabetes Association of the Republic of China — analyzing de-identified NHI data to perform risk classification for more than 2 million patients with Type 2 diabetes. Very few countries globally possess a single database of this scale and completeness that can be used to train medical AI.
Drones and Sensors in the Fields
At a fig orchard in Douliu, Yunlin, a farmer applied for a smart agriculture subsidy and implemented a smart skylight and smart irrigation system. Result: loss rates dropped below 5%.
This kind of story is repeating itself across Taiwan. The Council of Agriculture's "Grain Industry Upgrade Plan" has established 111 rice collective production zones nationwide, introducing sensors and AI image recognition technology to grasp soil and climate data in real time and determine optimal harvest timing. Farmers can use a mobile phone connected to the cloud to conduct a "field patrol" — no longer needing to walk a whole afternoon under scorching sun.
Drones are another rapidly penetrating entry point. Experimental data show that drone agricultural spraying can reduce application time by 95% compared to human labor, and pesticide usage decreases by 60%. A farmer in Zhongpu, Chiayi, used a drone to spray a loofah field — less than five minutes to cover two tan (roughly 0.2 hectares). In a rice field comparison experiment, traditional hand-spraying pest control achieved 63% effectiveness; a drone reached 93%.
Chiayi County, Kaohsiung City, and other local governments provide smart agriculture subsidies covering smart sensing systems, environmental control systems, drones, and other items, with subsidies reaching up to 50% of equipment costs up to a maximum of NT$500,000.
But the spread of technology also brings new problems. Disputes over drift from neighboring fields during drone spraying are frequent — a rice farmer in Chiayi commissioned a contracted spraying operator to apply herbicide, and the pesticide drifted to an adjacent tomato field, killing more than half the plants; the court ordered damages of more than NT$450,000. There are still not enough qualified drone pilots with the required dual certification; regulations require no spraying when wind speeds exceed 3 meters per second, and maintaining a 20-meter buffer from neighboring fields — but execution on the ground still falls short.
The technology is already there. Institutions and talent are still catching up.
The AI Tutor in the Classroom
The "AdaptLearn" (因材網) platform developed by the Ministry of Education has a built-in generative AI learning companion called "eDo" that can guide students through dialogue to clarify concepts, practice problems, and receive positive feedback. As of March 2025, more than 750,000 teachers and students were using it, with an average of more than 30,000 people online daily.
Behind these figures is the structural need of the 108 Curriculum Guidelines. The new curriculum emphasizes learning portfolio documentation, self-directed learning, and inquiry-based learning — all of which require individualized guidance, but Taiwan's teacher-to-student ratio does not allow every student to have a dedicated tutor. AI fills this gap.
Generative AI is also changing how teachers prepare lessons. Teachers are no longer constrained by textbook frameworks; they can quickly generate diverse teaching materials, questions, and learning pathway plans. In the cross-disciplinary competency teaching emphasized by the 108 curriculum, AI has become a teacher's research assistant — quickly integrating knowledge from different fields and generating initial lesson plans.
Civil institutions such as the Junyi Academy are also actively developing AI-assisted learning tools, providing adaptive learning pathways for elementary and middle school students. At the higher education level, the Ministry of Education is advancing the "AI Technology and Applications Talent Cultivation Plan" (AITCP), building cross-university AI curriculum maps.
However, the real challenge is not technological. When 46% of Taiwanese consumers have already used generative AI tools — and the 18-to-25-year-old cohort as high as 69% (MIC, 2024 Q4 survey) — what schools need to teach is not just "how to use AI," but "how to judge whether what AI produces is correct."
How Taiwanese People See AI
According to MIC surveys, more than 90% of Taiwanese people have heard of AI, and the proportion who consider themselves to understand generative AI rose from 24% in 2023 to 40% in 2024. Close to 70% of consumers believe generative AI has development potential.
But there is a gap between "having heard of it" and "trusting it." Taiwanese people's two major anxieties about AI are over-reliance on technology and privacy. Government regulation, data transparency, and sufficient background knowledge are the three prerequisites consumers require before using AI products.
Enterprise figures are even more telling. KPMG's 2025 Taiwan Industry AI Survey shows that the enterprise AI adoption index averaged only 32 points (out of 100), failing to break 40 for three consecutive years. Nearly 50% of enterprises have not announced an AI development strategy; the strategy score for the retail, trade, and services sector is only 13.7 points. More than 40% of enterprises report having no AI data governance measures.
In other words: Taiwanese consumers are already encountering AI extensively in daily life (LINE customer service, convenience store fresh food, NHI app), but most enterprises are still at the "tool assistance" level, lacking a complete AI adoption pathway. AI's penetration is being poured in from the consumer end — it is not being pushed down from corporate strategy.
In December 2025, the Legislative Yuan passed the Artificial Intelligence Basic Act in its third reading — 20 articles in full, specifying seven governance principles (sustainable development, human autonomy, privacy protection, information security, transparency, fairness, accountability), with the National Science and Technology Council as the central competent authority and the Ministry of Digital Affairs responsible for the risk classification framework. This is Taiwan's first dedicated AI statute, and one of the few in the Asia region to address AI governance at the level of a basic law.
Taiwanese attitudes toward AI are more open than Europeans, but also more anxious. The openness comes from LINE, NHI, and convenience stores already making AI part of daily life; the anxiety comes from no one telling them what "automatic system processing" actually decides — or how.
An AI Island That Grew from LINE Group Chats
Taiwan's AI daily life has a distinctive structure: it diffuses along existing life infrastructure, growing upward from below. LINE is communication infrastructure, convenience stores are retail infrastructure, national health insurance is healthcare infrastructure. AI is grafted onto systems Taiwanese people already trust — so penetration is fast, and it runs deep.
National investment is increasing. The Executive Yuan's "New Ten Major AI Construction Projects" plan expects to invest at least NT$190 billion (approximately US$5.9 billion) over four years, covering semiconductors and computing infrastructure as well as AI applications in healthcare, agriculture, education, and other areas of daily life. The National Science and Technology Council's "TAIDE" large language model for Traditional Chinese has already been applied in seven areas — including the agricultural knowledge retrieval system "Shen Nong TAIDE" and Taiwanese language education in K-12 schools.
But numbers tell only half the story. What will truly determine the future direction of AI in Taiwan is the store manager pressing confirm at 6:30 in the morning at FamilyMart, the family medicine doctor in a clinic reading AI risk reports, the farmer in Yunlin conducting a field patrol by mobile phone.
They won't say they are "using AI." They will say: "The system recommended it." "The app said so." "The machine calculated it."
AI in Taiwan is no longer the subject of technology news. It is part of the everyday vocabulary — it just hasn't been named that way yet.
References
- FamilyMart "Three-Reduction Policy" AI System Helps Reduce 370 Metric Tons of Food Waste Monthly — Foodnext, 2022
- NHIA Partners with Google: How AI-on-DM Diabetes Precision Health Education Is Rewriting Care for 3.2 Million Patients — tenten, 2026
- How Google and Taiwan Are Building an AI Blueprint for Public Health — Google Blog, 2026
- MIC Survey: 46% of Consumers Have Used Generative AI Tools — Market Intelligence & Consulting Institute, 2025
- LINE BIZ CONVERGE 2025: Official Accounts in Taiwan Exceed 3.09 Million — LINE Taiwan, 2024
- LINE Taiwan's Vision for an AI Agent Era — LY Corporation, 2026
- aetherAI, with NTU Hospital and Cathay General, Receives First Domestic AI Colonoscopy Polyp Detection Medical Device Certificate — Global Biotech Monthly
- A New Era of AI Governance in Taiwan: Industry Insights After the AI Basic Act Passage — AI Academy Taiwan, 2025
- KPMG Taiwan Industry AI Application Trends and Outlook Report — KPMG, 2025
- How Effective Is Drone Agricultural Spraying? Farmers Amazed as 2 Tan of Loofah Field Done in Under 5 Minutes — Newsmarket
- Ministry of Education AdaptLearn AI Learning Companion: Over 750,000 Teachers and Students Using It — Youth Daily News, 2025
Further Reading
- AI Industry (zh only)
- AI Development (zh only)