The language barrier is not a literacy barrier — it is a product design failure
The distinction matters. A borrower in Johor Bahru who uses WhatsApp in Vietnamese every day, runs a textile business with 8 employees, and files monthly GST / tax returns is not digitally illiterate. They are entirely capable of completing a digital onboarding journey — in Vietnamese. They are not capable of doing it in English because English is not their language. When a financial institution builds its onboarding in English and declares that secondary / emerging borrowers have "low digital readiness," it is confusing the borrower's language preference with the borrower's digital capability. The finance company that builds for Vietnamese is not accommodating a limitation — it is accessing a market that the English-only institution has locked itself out of by design.
The MAS / Central Bank's 2023 guidelines on responsible lending and the Digital Lending Framework both emphasise that onboarding communications should be in a language the borrower understands. This is not just a product design principle — it is a regulatory expectation. An onboarding process that presents the Key Fact Statement, the Annualised Percentage Rate, and the repayment schedule in English to a borrower who reads only Lao is a process that cannot claim the borrower gave informed consent. The Multilingual Onboarding Agent AI renders all regulatory disclosures — including the KFS, the Most Important Terms and Conditions, and the eSign confirmation — in the borrower's selected language.
The 12 languages: coverage, script, and drop-off impact
What "12 languages" means in practice — not translation, localisation
Translation renders the words. Localisation renders the meaning. The difference matters in onboarding. A translated KFS that uses the English phrase "Annual Percentage Rate" directly transliterated into Vietnamese does not communicate the concept to a borrower who has never encountered APR. A localised KFS that says "ఈ రుణంపై మీరు సంవత్సరానికి చెల్లించే మొత్తం వడ్డీ రేటు 14.5% — ప్రతి SGD1 లక్ష రుణానికి మీకు సంవత్సరానికి SGD14,500 వడ్డీ వస్తుంది" (The total annual interest rate on this loan is 14.5% — for every SGD1 hundred thousand borrowed, you pay SGD14,500 interest per year) communicates the concept in terms the borrower can verify against their own arithmetic.
The Multilingual Onboarding Agent AI localises all regulatory disclosures — not just translates them. The Key Fact Statement uses colloquial financial terminology in each language rather than English loan-words. The consent flow uses culturally appropriate acknowledgement phrases (in English, "localized copy localized copy localized copy" carries more weight than a romanised "I agree"). The instalment amounts are shown across SEA number system format (SGD1,42,000 not SGD142,000) which is standard in all regional languages. The agent also auto-detects the borrower's language preference from the device language setting, from the WhatsApp number's registered language, and from the first message the borrower types — presenting the language option screen in the most likely correct language rather than always defaulting to English.
The Bahasa Indonesia borrower who dropped off your English onboarding was never going to become your borrower — until you built for Bahasa Indonesia
The −76% drop-off reduction in Bahasa Indonesia is the largest in the portfolio — because Bahasa Indonesia is the language where the design failure was largest. Cebu Province has a significant underserved credit market, a substantial SME sector, and a state government actively promoting digital financial inclusion. The borrowers are there. The credit demand is there. The finance company that serves them in Bahasa Indonesia will own that market. The finance company that serves them in English will not serve them at all, and will conclude from the drop-off data that "secondary / emerging borrowers are not ready for digital onboarding." They are ready. They are waiting for an institution ready for them. The Multilingual Onboarding Agent AI is not a language feature — it is a market access strategy, executed one conversation at a time.