Southeast Asia is going paperless on a deadline. Vietnam's Circular 13/2025/TT-BYT makes electronic medical records mandatory and bans paper charts from 2027. The records are arriving. The clinical AI meant to sit on top of them was built for American hospitals, English content, and Western formularies — and most of it was never built for this region at all.
That gap is the whole opportunity, and it is wider than the marketing suggests.
The mandate is real, and it has dates
Circular 13/2025 is not a recommendation. It requires every hospital in Vietnam to run electronic records as the primary chart, eliminate paper in stages through 2027, and reach full coverage by 2030. Hospitals are moving: 881 of 1,645 had announced electronic-record implementation by early October 2025, up from 142 in late March. A six-fold jump in two quarters.
Vietnam is not alone. Indonesia reports 96% of hospitals on electronic records and 92% connected to its national SATUSEHAT platform. The charts are going digital across the region, fast and by law.
That switch costs real money. An average Vietnamese provincial hospital needs more than VND 10 billion — about USD 383,000 — to stand one up, and cost is the leading reason the holdouts hold out. After all of it, the hospital has a digital record. It does not yet have a second opinion.
A digital chart is not a clinical decision
An electronic record makes the chart searchable. It does not make it safer to prescribe against. That distinction is where patients get hurt.
In two Vietnamese provincial hospitals, a peer-reviewed study found a medication-error rate of 39.1% across 5,271 doses, with 92% of patients hit by at least one error (Nguyen et al., 2015). The contributing factors named in study after study across the region are the same two: staff shortage and workload. Vietnam has about 12.5 doctors per 10,000 people; Indonesia and Thailand have fewer than half as many. The doctors who are there carry more patients than a careful chart review survives.
This is the actual job for clinical AI here. Not to store the record — the mandate already forces that — but to read it the way an overworked clinician cannot: catch the interaction, question the dose, summarize a fragmented history before the next patient is already in the room. The mandate built the shelf. Something still has to be the second set of eyes.
Built somewhere else
The obvious move is to buy what the West already built. It does not fit, and it is worth being precise about why, because the mismatch hides behind a localization label.
UpToDate is the reference most Asian clinicians know. It supports search and navigation in 16 languages, Vietnamese and Bahasa Indonesia among them — which is easy to mistake for localization. It is not. The search box speaks Vietnamese; the clinical content stays in English, on Western guidelines and formularies, with no national drug list and no insurance-code mapping. And the AI-enhanced edition Wolters Kluwer launched across APAC in 2025 covers Australia, New Zealand, Singapore, Taiwan, and Hong Kong. Not Vietnam, Thailand, Indonesia, Malaysia, or the Philippines — the five largest markets in the region, left off the map.
OpenEvidence, the AI tool clinicians keep comparing us to, gates access to verified US clinicians with a National Provider Identifier and pulled out of the EU and UK in 2026 over regulatory uncertainty. A doctor in Hanoi cannot log in.
The honest part: there is a serious Southeast-Asia-native player. MIMS covers 17 Asia-Pacific markets and millions of clinicians. But MIMS is a drug-reference and interaction checker — it tells you whether two drugs clash, not how to reason through the patient in front of you. A different category, and pretending the region is empty would be exactly the kind of claim this argument is against.
Localization is not translation
A clinical reasoning engine for this region has to know the national formulary a Vietnamese doctor actually prescribes from, the local treatment guidelines a Western model never trained on, the insurance codes that decide whether a prescription gets reimbursed, and the language the patient describes their symptoms in — which, as we wrote about before, carries clinical signal that does not survive the trip through English.
That is what we build Meddies to be. Not a translated reference and not a drug-checker, but the clinical-reasoning layer that sits beside the doctor, inside the electronic record the mandate just made compulsory. It reads the chart, screens the prescription, and answers with its sources shown — against the formulary, the guidelines, and in the language of the region it serves.
The region didn't import its medicine
Southeast Asia did not import its doctors, its hospitals, or its drug formularies. It built them, for the conditions, the budgets, and the patients it actually has. The policy that just forced electronic records is regional, written for regional medicine.
The clinical AI that reads those records should be built the same way — for the region, not translated into it after the fact. That is the bet. The deadline to make it is already on the calendar.