Shivani Siroya built Tala on the lives banks could not read.

Ask a room of Kenyans to name three digital lenders and Tala will rarely be missing. The brand has occupied Kenyan phones for more than a decade, long enough to feel indigenous. Its loans arrive through M-PESA, its advertisements speak in the language of daily Kenyan life, and its legal operator sits in Nairobi.

Tala is American in corporate origin. It is headquartered in Santa Monica, California. Its Kenyan business is Inventure Mobile Limited, trading as Tala, which the Central Bank of Kenya licensed as a digital credit provider on 30 January 2023. Kenya was never a peripheral branch on Tala’s map, however. It was the market in which Shivani Siroya tested whether a phone could make an economically active person visible to a financial system that had no useful record of her.

The experiment has become a global fintech company. Tala’s current homepage reports more than 15 million customers served, more than one billion proprietary data signals and over US$10 billion in credit access delivered. Those are cumulative company figures. A narrower independent measure reported by Forbes in September 2025 put Tala’s active revenue-generating customers at 1.8 million. Earlier expansion included Tanzania, but Tala’s present company map lists Kenya as its East African location alongside businesses and partnerships across Asia and Latin America. Even on the narrower customer measure, the small lending experiment that began in Kenya has travelled a considerable distance.

Before the algorithm, there was Seema.

Siroya was born in India and grew up between India and the United States. Her mother practised medicine in Brooklyn and sometimes allowed patients to pay later, a memory Siroya would later recount in her TED talk on credit and identity. Credit, in that household, was not an abstract score. It was a judgment about a person, made under uncertainty, followed by a relationship.

Siroya studied government and international relations at Wesleyan University, then completed a master’s degree in public health at Columbia University, specializing in health economics and international policy. Her early career took her through healthcare equity research at UBS, health-costing and economic-development work with the United Nations Population Fund, and later mergers, acquisitions, and corporate strategy involving Citigroup and Health Net.

The job at the UN took her across markets in Africa and Asia. She interviewed thousands of small-business owners about income, stock, household obligations, and the ways they managed when money arrived irregularly. The formal system often saw no payslip, collateral file, or bureau history. Siroya saw businesses, customers, school fees, savings habits, and people making disciplined decisions with very little room for error.

When some of the entrepreneurs could not obtain working capital, she began lending from her savings.

One of them was Seema, a woman in Chennai who ran a tile-making business and a jewelry shop and was raising two sons. Siroya made several small loans to her. Seema repaid each one on time.

A visiting friend wanted to know why Siroya trusted her. Siroya knew that customers gathered at Seema’s shop. She knew Seema saved about 30 percent of her income towards one son’s computer science studies and helped younger women in her community plan weddings. None of it appeared in a conventional credit file. Together, those details described commercial activity, obligations, standing, and intent.

Seema also knew who had lent to her. Repayment protected a relationship she might need again.

Siroya later described it as “a financial relationship that I wanted to figure out how to replicate at scale." Tala grew from that exchange. The first problem was not a shortage of willing borrowers. It was a financial system unable to read the lives they were already living.

Building a lender while keeping the day job.

Siroya did not leave employment with a finished app or a venture-capital term sheet. She contacted about 1,500 people through LinkedIn while looking for anyone solving the same problem. She learned enough code to build an early prototype and used her savings to test it. A mentor advised her to keep six to twelve months of rent in the bank before resigning, so she continued working full-time while the idea took shape.

The company founded in 2011 was called InVenture. Its early products included a text-based record-keeping tool named InSight, intended to help small businesses maintain simple accounts and develop evidence of creditworthiness. The original plan was to produce scores that established lenders could use. Banks remained slow to lend against them.

InVenture then chose to risk its own capital on its own conclusions. Its Android lending app, Mkopo Rahisi, launched in Kenya in March 2014. M-Shwari had already introduced digital credit to Kenya in 2012, so Tala did not invent the country’s first digital loan. It was an early app-based lender to use smartphone information to decide which unsecured borrowers qualified and on what terms, often within minutes. Mkopo Rahisi became Tala in 2016.

A conventional lender asked the customer to prove that she resembled people it had served before. Tala built a system around people the conventional file had excluded. The phone became an application form, a source of behavioral information, a decision engine, and a route to M-PESA.

Capital followed. Tala’s reported financings rose from a US$1.2 million seed round in 2013 to a US$10 million Series A in 2016, US$30 million Series B in 2017, US$65 million in equity and lending capital in 2018, US$110 million in Series D financing in 2019 and a US$145 million Series E led by Upstart in 2021. The company now says it is backed by more than US$500 million in funding, a figure that combines investment and debt rather than describing a single equity valuation.

The fellowships and honours accumulated alongside the capital. Siroya became an Echoing Green Fellow in 2011, a TED Fellow in 2013, an Ashoka Fellow, an Aspen Finance Leaders Fellow and a World Economic Forum Young Global Leader. When WIRED marked its twenty-fifth anniversary, Melinda Gates selected Siroya as one of the people she expected to shape technology’s next twenty-five years.

Access can change a life and still require scrutiny.

Tala’s Kenyan product now offers a reusable credit line of up to KSh50,000, paid into M-PESA. A customer may choose a repayment date of up to 61 days and pay early without a penalty. Tala publishes daily interest of 0.3 to 0.6 percent, equivalent to an annual percentage rate of 109.5 to 219 percent before any late charge. At those daily rates, KSh10,000 held for 30 days attracts KSh900 to KSh1,800 in interest.

Speed removes the journey to a branch, the collateral demand and the embarrassment of asking a relative. Price determines how much of that freedom survives repayment.

A 2022 study conducted by 60 Decibels for Tala interviewed 250 randomly selected Kenyan customers. Forty-four percent said Tala had given them their first access to a digital loan, 70 percent used the money for business and 84 percent reported an improvement in their quality of life. Forty-seven per cent were borrowing from Tala every month. Three-quarters used savings to repay, while 29 percent had borrowed elsewhere to meet a Tala repayment. The sample described Tala’s customers rather than Kenya as a whole and was somewhat better off than the national population.

A peer-reviewed study published in The Accounting Review examined an earlier Tala experiment. Applicants close to the approval threshold were randomly granted or denied loans averaging about KSh3,675 for roughly 28 days. Approval raised the reported likelihood of employment or self-employment by 24 percent and self-reported monthly income by 21 percent. The results apply to marginal applicants in that historical experiment, not every borrower or every digital lender. For that group, a small and costly loan created measurable economic opportunity where no loan had been available.

The 2024 FinAccess Household Survey found that formal financial access had risen to 84.8 percent, yet only 18.3 per cent of adults were financially healthy. Among borrowers, 16.6 percent had completely defaulted on a loan, up from 10.7 percent in 2021. Access has expanded faster than resilience.

FinAccess measures the entire credit market rather than Tala’s portfolio. Tala’s own evidence places repeat borrowing at the centre of its model, while its Kenyan website advertises continuous access and rising limits after timely repayment. A useful credit line helps a trader bridge stock and sales. A costly credit line used every month may also reveal that income and expenses never quite meet.

Tala should publish how pricing, repayment periods and financial health change for the same groups of customers over time. A borrower who repays repeatedly should become cheaper to serve, qualify for longer and more suitable finance, build savings, or carry a recognised positive history to another provider. Disbursement records activity. Progress records inclusion.

The life inside the data.

The story that Siroya once learnt by sitting with Seema is now assembled by software. Tala’s Kenyan privacy material lists identity and transaction information alongside device identifiers, location and certain SMS metadata or keywords used for fraud checks and lending decisions. A separate channel notice also lists categories such as installed applications, contact lists and call or SMS logs, depending on the channel and permissions used. Tala says it does not call a customer’s friends or relatives for collection, does not sell personal details to marketers and allows a customer to request human reconsideration of an automated decision.

Kenyan law now places boundaries around the practice. Digital lenders must disclose total cost and annual percentage rate, assess a borrower’s ability to repay, collect only data reasonably required for their function, protect personal information, resolve complaints and avoid abusive collection. Inventure Mobile appears in the Central Bank’s July 2026 directory among 252 licensed digital credit providers.

Licensing established a floor. Tala’s ambition demands a higher ceiling. A model built from intimate behaviour should tell a customer which broad factors affected an offer, give her a practical route to correct bad information and retain no more data than the decision requires. Tala’s current terms say it reports positive and negative repayment information to credit-reference bureaus, creating a route beyond its private dataset. The next measure is whether that positive record unlocks better outside offers and whether a borrower can easily inspect and correct it. Otherwise the same life that was invisible to a bank becomes visible to one private system and remains of little practical value everywhere else.

From lending app to financial infrastructure.

By September 2026, Siroya is building beyond the original direct-to-consumer app. Tala describes itself as “AI-native credit infrastructure for the global majority”. The phrase refers to people whose economic lives are active but incompletely recorded by formal finance. AI-native means that machine learning and alternative data sit inside the credit decision from the beginning rather than being added to an old bank-scoring process. The shift turns Tala’s decision system into a product that banks, wallets and capital providers can use to reach customers through their own channels.

The latest independent account of Tala’s economics came from Forbes in September 2025. It reported revenue running at about US$340 million a year, growth of 35 percent and a continuing loss, with management targeting break-even in early 2026. Tala had not announced whether it met that target by 6 September 2026. The transition to infrastructure is therefore also a search for scale that does not depend entirely on acquiring and funding each borrower through Tala’s own balance sheet.

In January 2026, Tala announced a US$100 million partnership with CIMB for Vietnam. Tala supplies the technology while the licensed bank provides the loans. The product offers eligible customers a line of up to VND30 million, with each draw running for as long as 61 days.

In Guatemala, an integration announced in May places Tala-underwritten credit inside the Airtm wallet and disburses it in USDC, a digital token designed to track the US dollar. A US$50 million facility announced with Huma Finance and Solana in December 2025 is intended to finance tokenized loans. The structure turns expected loan repayments into digital assets that investors can fund over blockchain-based rails.

Tala’s decision engine can now travel into a wallet or bank without Tala acquiring every customer through its own app. Global investors gain exposure to repayments from markets they would struggle to assess directly. Borrowers gain another route to liquidity, but the currency of the debt, conversion costs, wallet access, data sharing and responsibility for complaints must remain legible at the customer’s end of the transaction.

Tala is also placing artificial intelligence at the centre of the credit decision. On 31 August 2026, it said it was working with AI research company Embed on a ledger foundation model. The system is intended to learn from long sequences of transactions and their surrounding context, including repayment, default, app engagement and customer-service interactions. Public blockchain records would provide an initial body of training material. Tala then proposes to test the learning on its own records and build artificial datasets that preserve useful patterns without exposing individual identities.

Tala acknowledges that behaviour on a public blockchain may not translate cleanly to consumer lending. Bots, speculative trading and the structure of decentralized finance differ from the irregular earnings and household choices of a borrower in Nairobi or Chennai. Model performance will need to be tested across gender, income, language, location and device type. Privacy protection must preserve whole sequences without allowing an artificial record to be traced back to a real person.

Siroya’s first insight survived every change of technology. Context can reveal capacity that a formal score misses. The power attached to that insight has changed. In Chennai, one woman chose what to tell one lender. At global scale, a lender, a wallet and a capital provider may jointly infer far more than a customer knowingly offered.

Trust must run in both directions.

Seema repaid because trust ran both ways. Siroya knew the life behind the request, and Seema knew the person who might be there when capital was needed again. Tala has spent fifteen years converting the first half of that relationship into software. It can recognise a borrower at a scale no human lender could approach.

Its next achievement should belong to the customer. Every sound repayment ought to reduce uncertainty, improve terms and build a financial identity she can inspect, correct and carry. Tala’s proposed portable digital reputation points in that direction. Portability should become a delivered customer right rather than a promise attached to new financing rails.

Kenya gave Tala the market in which it could prove that informal lives were not unbankable. Tala now has the capital, data and distribution partners to prove something harder. A system that learns from every repayment but leaves the borrower dependent on the same short-term lender has improved underwriting without completing inclusion.

Seema’s story did not ask technology to know everything about her. It asked finance to recognise enough of her life to say yes, then allow that trust to compound in her favour.