On January 5, 2013, Nguyen Nguyen sat in a bedroom overlooking the Hudson River and began to code. His daughter, Sophie, had been born three days earlier. The screen held the first version of a Facebook-based credit score; the larger idea was still almost impolite in its ambition. A person without a conventional credit file did not necessarily lack creditworthiness. Perhaps the file was the thing that was incomplete.
Nguyen understood the file from the inside. Born and raised in Vietnam, he came to the United States in 2000, trained in game theory and econometrics, and completed a doctorate at Rice University. He later worked in Barclays' global credit-risk operation in New York, building models for credit and fraud decisions across portfolios. There, a recurring institutional problem became visible: in markets where credit-bureau coverage was thin, banks could not confidently assess large parts of the population. A missing record behaved like a closed door.
The insight behind Trusting Social was to look for other handles. Telecom activity, consumer behavior and other alternative data could supplement conventional bureau information. Machine learning could find patterns at a scale no loan officer could match. The proposition was not that everybody should receive a loan. It was that a lender should be able to distinguish a bad history from no history at all.
That distinction carries an economic weight larger than the score itself. When a bank cannot separate an unfamiliar customer from a risky one, it can decline both, demand more paperwork or price uncertainty into the loan. The cautious choice is understandable at the level of one institution. Repeated across a market, it becomes an invisible tax on anyone whose financial life does not resemble the data used to train the old system. Nguyen chose to work on that translation layer: turn the ordinary traces of economic life into evidence a regulated lender could use. The job required more than clever modeling. It required integrations with institutions that move carefully, distribution partners that reach mass-market customers, and rules for handling data that belongs to real people. The startup's technical thesis and its social mission were therefore bound to the same operational question: could a new signal survive contact with a bank?
“I was driven by a single vision: to give every person in the world access to credit.”Nguyen Nguyen, reflecting on the company's beginning
The file was never the person
Traditional credit scoring works well when its inputs exist: loans, repayments, balances, defaults, years of financial behavior gathered into a bureau. The machinery becomes less useful when somebody earns, shops and pays bills outside the formal channels that feed it. In fast-growing Asian markets, that gap can turn ordinary consumers into statistical blank spaces. Banks face a rational constraint; people experience it as rejection, paperwork or an offer priced too high.
Trusting Social was built between those perspectives. It sells infrastructure to banks and finance companies rather than replacing them. Its credit insights help institutions estimate risk; its acquisition tools help find likely customers; its identity products verify that an applicant is real. The company says its risk profiles cover more than one billion consumers across India, Indonesia, Vietnam and the Philippines. In a 2021 financing announcement, it said more than 130 financial institutions used its products.
Scale is the seductive number here, but the more revealing unit is the decision. A score has consequences only after a lender chooses a limit, a price and an explanation. Nguyen's public argument has gradually moved from access toward cost. At an IFC conference in 2025, he reflected that Vietnam had made substantial progress on credit access; the next task was a richer data ecosystem that could lower what borrowers pay. Inclusion, in this telling, is not complete when the application form becomes available. It is complete only when the product is suitable and affordable.
A score grows limbs
The company's product history reads like a tour of everything that can go wrong after a model says yes. An applicant still has to prove an identity. A document may be worn or photographed badly. A fraudster may present a replayed face. A legitimate customer may abandon a slow process. Trusting Social launched an eKYC product in 2020, using optical character recognition, face matching and liveness checks to move those moments onto a phone.
Then generative AI widened the aperture again. In 2023 the company introduced Agent Foundry, a platform for trainable enterprise agents. Its early banking agent, ALICE, was designed to personalize customer conversations. The company later moved deeper into automated collections in the United States, an area where the quality of a conversation matters precisely because the context can be difficult. The technological arc runs from prediction to verification to language. The responsibility expands with it.
Nguyen has summarized the philosophy in a compact line: “Personalized banking is a right and not a privilege.” The phrasing matters because personalization has often been sold as a luxury - a wealth manager's attention, a hand-tailored portfolio, a premium service tier. Software makes a different wager. If the marginal cost of a relevant offer or useful conversation falls far enough, attention can travel down-market.
But a more observant financial system is not automatically a fairer one. Alternative data introduces hard questions about consent, relevance and the ability to challenge a decision. Identity technology concentrates sensitive information. An autonomous agent can make a process humane or merely make it relentless. Trusting Social says it anonymizes data and emphasizes privacy and regulatory compliance. Those commitments are not accessories to the model. They are part of whether the model deserves to operate.
The global village and the long week
Nguyen's own telling remains unusually close to the keyboard. In a 2026 reflection, he described thirteen years and more than 50,000 hours of work, still running at 80 hours a week. He contrasted the solitary bedroom prototype with a “global village” of roughly 400 people. The figures are his, offered less as a victory lap than as an accounting of interdependence. Every client promise, he wrote, eventually touches the people building and supporting the system.
He named his wife Ha Nguyen, co-founder and chief scientist Tuyen Huynh, and executives Jaideep Lakshminarayanan, Siddhartha Kohli and Johnny Escaler. Tuyen, a machine-learning PhD, has been Nguyen's collaborator for two decades. That continuity is easy to miss beneath product announcements. Trusting Social combines the vocabulary of a research lab, a regulated vendor and a consumer-finance operator. It requires people who can translate among all three.
The company's partnerships reveal the same hybrid. Banks supply balance sheets and regulatory relationships. Telecom and consumer platforms supply distribution and, under governed arrangements, new signals. In 2022, Masan Group's subsidiary The Sherpa Company announced a $65 million investment tied to a plan for tailored fintech services across Masan's consumer ecosystem. In Vietnam, the pitch put scoring technology beside everyday retail rather than inside a distant bank branch.
“Personalized banking is a right and not a privilege.”Nguyen Nguyen on the purpose of lower-cost AI
By June 2026, Trusting Social was describing a partnership with electronics retailer Dien May Xanh that would place its EVO Money digital loan experience across more than 3,000 stores. The old abstraction - a credit score for someone without a conventional file - had become a very concrete scene: a shopper, a phone, an identity check, a risk decision and a loan offer delivered while the purchase was still in view.
The next file is a conversation
Nguyen began by asking whether new data could reveal creditworthiness. The arrival of capable language models has turned that question outward. Can an automated system explain a choice, understand intent and remember context across channels? Can it lower operating costs without making the customer feel processed? He has become an energetic advocate for AI agents while also speaking publicly about the social disruption that advanced AI could bring.
That combination - acceleration and unease - feels appropriate. Credit is a domain where technical optimism meets the sharp edge of consequence. A slightly better prediction can open a useful line of credit; a bad decision can burden a household. A fast identity check can eliminate a trip to a branch; a careless one can expose private data. An agent can answer at midnight; it can also misunderstand at exactly the wrong moment.
The company that began with a Facebook score now sits inside that tension. Its most consequential work is not making finance look futuristic. It is making a succession of ordinary decisions more accurate, less expensive and easier to complete. Nguyen's original insight still sets the standard: absence of evidence should not be confused with evidence of absence. The person was always there. The financial system needed a better way to look.