Trusting Social

Here is a problem that sounds like a riddle and is actually a business plan. A bank wants to lend money to someone. To decide whether that is a good idea, it pulls the person's credit history. But the person has never borrowed money before, so there is no history to pull. Which means the bank can't tell if they'll repay. Which means the bank won't lend. Which means the person never builds a history. The circle closes, and roughly a billion people in Asia stand outside it, "financially invisible," as the industry politely puts it, which is a nicer way of saying the machinery was never built to look at them.

Trusting Social's entire proposition is that the machinery can be built. Founded in 2013 by Nguyen Nguyen, a former Barclays credit-risk quant with a PhD in econometrics, alongside machine-learning PhD Tuyen Huynh and co-founder Rohit Narang, the company decided the missing credit file wasn't actually missing. It was scattered - across telco records, web behavior, and mobile signals - and if you had enough data science, you could assemble it into something a lender would trust.

The word "trust" is doing a lot of work in the name, and it should. Lending is, at bottom, a bet on trust, and traditional credit scoring is just a formalized way of saying: we trust you because a bureau vouched for you. In markets where the bureau is thin or nonexistent, Trusting Social offers a substitute vouching mechanism. Its flagship product, Trust Scores, uses machine-learning models trained on alternative data to produce a credit risk score for someone who has, on paper, no risk profile at all.

The picks-and-shovels move

The clever structural decision - and it is genuinely clever - is that Trusting Social does not lend money. Lending is capital-intensive, regulated to the eyeballs, and exposes you to the actual credit losses. Instead, the company sells the scores, the identity checks, and the customer targeting to the banks that do lend. Every lender becomes a potential customer rather than a competitor. It is the toll booth on a road that other people paid to pave.

That B2B posture shows up across the product line. Trust Vision Solutions handles digital onboarding and eKYC - the "know your customer" identity verification that is legally mandatory and operationally miserable in fraud-heavy markets. It uses facial recognition, liveness detection, and document OCR to let a bank onboard a customer it has never physically met, in seconds rather than days. Trust Vision, in effect, is the answer to a second riddle: how do you trust a face you have never seen? You build a model that has seen a great many faces.

Then there is Smart Customer Acquisition, which flips the telescope around. Instead of scoring the borrowers who walk in the door, it helps lenders find the credit-qualified borrowers who never walked in at all - the invisible ones who would, in fact, repay. The company reports that this product has facilitated roughly $800 million in personal loans. That is the mission and the margin pointing the same direction, which is the only configuration in which a mission tends to survive contact with a P&L.

The scale of the thing

The numbers Trusting Social attaches to itself are the kind that only make sense at data-infrastructure scale. It says it covers more than a billion consumers with credit profiles, serves 130-plus financial institutions (a figure it has quoted as high as 170), and processes something like 120 billion records every week. The company operates across four Asian markets - Vietnam, India, Indonesia, and the Philippines - each with its own regulator, its own data, and its own definition of the underbanked.

Behind those numbers sits a workforce that reads more like a research lab than a fintech: about 320 employees, including a reported 20-plus PhD data scientists and 80-plus holders of a Master of Science. That density of advanced degrees is not decoration. Alternative-data credit scoring is a modeling problem before it is a software problem, and the hard part - separating signal that predicts repayment from signal that just correlates with being poor - is exactly the sort of thing you hire econometricians to worry about.

Enter the agents

In 2023 the company did the thing every data company eventually does when a new capability arrives: it pointed the capability at itself. Trusting Social launched Agent Foundry, a platform of generative-AI banking agents built on Microsoft Azure, becoming one of the earliest adopters of Azure OpenAI Service in Vietnam. The first agent, ALICE, handles sales and relationship management - onboarding, customer queries, conversational cross-sell. One client, the company says, saw a 10% lift in credit-card conversion after deploying it. Two more agents, ANANDA for analytics and ALAN for software, are in development.

It is worth noting what this move implies. A credit-scoring company deciding to build conversational agents is a company that has concluded the next bottleneck isn't data - it's the conversation. You can score a billion people, but if a bank can't actually talk to them at scale, the scores sit unused. The agents are the company's answer to its own success.

The money, and the accounting

Trusting Social has raised over $200 million across its life, drawing early checks from Sequoia Capital India, BeeNext, and 500 Global. The defining round came in April 2022, when The Sherpa Company - a subsidiary of the Vietnamese conglomerate Masan Group - led a $65 million Series C and took a stake reported at around 25%, the first tranche of a planned $105 million strategic investment. That is a specific kind of capital: your investor is also your distribution, with a retail footprint that gives the scores somewhere to go.

The financials are a reminder that inclusion is a long game. The Singapore-registered parent, Trust IQ, reported revenue of about $23.1 million in 2025, up modestly from $21.2 million the year before - and an $87.5 million net loss, driven largely by a fair-value revaluation of financial liabilities rather than the operating business. That is an accounting event more than a cash event, though it is the sort of line that reminds you a mission still has to clear an auditor.

Who actually uses this

The customer for all of this is not, in the first instance, the borrower. It is the bank, the consumer lender, the telco - the institution that has the license and the balance sheet and the very real fear of lending into a fraud it can't detect. Trusting Social sells that institution three things it struggles to build in-house: a way to price risk on a thin file, a way to prove an applicant is a real person, and a way to find good customers it would otherwise never reach. The borrower benefits downstream, when a loan gets approved that a legacy scorecard would have rejected on the grounds that the applicant did not, statistically speaking, exist.

That framing matters because "financial inclusion" can slide into sounding like philanthropy, and Trusting Social is not a charity. It is a data business with a mission that happens to be commercially load-bearing. The unbanked are underserved precisely because they are difficult - hard to score, hard to verify, hard to reach - and difficulty is where a defensible business lives. Anyone can lend to a customer with a pristine bureau file. Building the model that lets a bank lend to the customer without one is the part competitors can't trivially copy, which is also why the field is crowded but not commoditized: CredoLab, LenddoEFL, Tala and others are all chasing versions of the same hard problem.

There is also the matter of trust flowing the other way. A company that ingests telco, web, and mobile data to decide who gets credit is holding a great deal of sensitive information about people who never asked to be scored, and the regulatory temperature around exactly that practice has been rising across every market it operates in. Trusting Social's public posture leans on anonymization, on-premise processing options, and compliance with local data-protection rules - which is less a marketing flourish than a precondition for the whole model to keep operating. The permission to score a billion people is not permanent; it is renewed, market by market, by not abusing it.

In 2026 the company signed on to distribute its EVO Money product across more than 3,000 Dien May Xanh retail stores in Vietnam, which is what building the file, at scale, every week, eventually looks like when it reaches a checkout counter. A customer buying an appliance can be scored, verified, and offered credit on the spot - the entire apparatus of econometrics and machine learning and facial recognition collapsing into a single moment at a till. Whether that is empowerment or just faster access to debt is a fair question, and it is the question that hangs over the entire category. Trusting Social's answer, implicit in everything it builds, is that the credit was always going to be extended somewhere - better it be priced by a model that can actually see you than denied by a bureau that never could.