The code Kamal Budhabhatti came back to finish.

The bank was waiting for its software. Kamal Budhabhatti was in India because Kenya had deported him before he could finish writing it.

He had arrived in Nairobi from Jamnagar, a trading city on India’s western coast. His father sold newspapers. His mother raised three children with little money and a practical genius for making scarce things last. Budhabhatti studied physics, learnt computer programming in his spare time and took a data-entry job in Kenya after a friend told him about the opening.

The work placed him behind a computer but gave him little to create. He spent his days transferring information from old files into a machine. Then a friend mentioned a bank that needed clearing-house software.

Budhabhatti began building it while he was still employed. His employer discovered that he had used company time and equipment for the outside project and dismissed him. According to a 2011 Forbes account, he was then deported.

The program remained unfinished. The bank remained interested.

Budhabhatti raised the fare back to Nairobi. He stayed in a friend’s small apartment, worked late into the night and has recalled eating once a day while he completed the system. The bank accepted it and paid him. Other financial institutions brought him more work.

Craft Silicon grew from the program he returned to finish.

The company now reports more than 1,200 employees, over 350 financial-institution clients in more than 30 countries and systems that reach over 150 million end users. Its founder remains Group Chief Executive, but he no longer runs the African operating company from day to day. He is trying to complete a transfer that is harder than returning to Nairobi with no money.

The programmer has to remove himself from the program.

The bank beneath the branch.

Craft Silicon dates its establishment to 2000. Bankers Realm became its early breakthrough, taking Budhabhatti from one clearing-house assignment into the systems that hold a financial institution together.

A core-banking system is the bank beneath the branch, the app and the agent. It records deposits, loans, interest, repayments and balances. A teller depends on it to know what belongs to the customer standing at the counter. A credit officer needs it to see what a borrower owes. The finance team needs it to close the books.

Replacing such a system is expensive and dangerous. Data must move accurately, staff must learn new processes and customers must retain access while the institution changes the machinery underneath them. A vendor that wins the installation assumes years of support and the possibility that one defect can interrupt thousands of financial relationships.

Equity Building Society became one of Craft Silicon’s formative customers. A Stanford Graduate School of Business case study records that Equity still processed its banking services manually before installing Bankers Realm as its first core system in 2000. In a 2024 interview with Frontier Fintech, Budhabhatti described the relationship with James Mwangi as an exchange between banker and programmer. Mwangi taught him how bank ledgers worked; Budhabhatti translated banking operations into software.

Bankers Realm helped move Equity from manual processing, but the institution’s growth soon tested the platform. By 2004, the Stanford case found, the system was hanging under the expanding volume of customers and could not accommodate the ATMs Equity wanted to introduce. Bankers Realm consultants were stationed almost permanently at the bank’s headquarters to keep it running.

Equity signed for Infosys’s Finacle system in June 2005. The new core went live in November and was deployed across all branches by the end of the year as part of an information-technology transformation valued at about US$8 million. Craft Silicon’s software had helped digitise a growing building society. The customer then grew beyond what that generation of the product could carry.

Craft Silicon could understand an African institution well enough to replace manual banking. It also had to increase capacity before the customer’s growth converted intimacy into dependence on engineers stationed beside a strained system.

Budhabhatti built for institutions that global software companies did not always understand. Microfinance lenders, savings and credit co-operatives and smaller banks in Africa and Asia needed products suited to local lending methods, mobile money, intermittent connectivity and balance sheets that could not absorb the cost of the largest international platforms.

Craft Silicon expanded into microfinance, payments, switching, digital channels and mobile banking. Its Bangalore operation opened in 2004. BR.NET followed in 2008 and moved more banking functions online. The company exported Nairobi-built financial infrastructure before fintech became a label broad enough to cover almost any application connected to money.

Forbes reported in 2011 that Craft Silicon had more than 200 clients in 40 countries, about 200 employees and annual revenue of roughly US$6 million. It placed the company’s value at about US$50 million, an estimate that was not tested through a public transaction.

Craft Silicon remains privately held and does not publish consolidated group accounts. The scale figures in its current public materials are therefore company-reported rather than the disclosures of a listed business. Named deployments across banks, microfinance institutions and SACCOs establish a substantial installed base. The harder comparison across years remains unavailable without revenue, profit, contract-renewal and regional-performance figures prepared on a consistent basis.

Customers entrust Craft Silicon with systems they are reluctant to replace. Longevity can be evidence of product reliability, switching cost or both. The next phase requires the company to show how much of its growth comes from new installations, expanded work for existing clients and newer products beyond core banking.

The founder once needed a bank to take a chance on unfinished code. The institutions now buying from Craft Silicon need evidence that the company supporting their most important systems will endure beyond its inventor.

The one-shilling salary.

Budhabhatti has never behaved like a founder content to disappear behind enterprise software.

In December 2013, he announced that his monthly salary would fall from seven figures to one Kenyan shilling. Craft Silicon was not in distress. Its board had approved a performance-linked arrangement under which bonuses would replace the large fixed salary.

The gesture made him one of the most memorable chief executives in Kenya. It also carried less personal danger than the number suggested. Budhabhatti owned part of the company and said the previous year’s bonus could support his lifestyle. One shilling changed the fixed line on his payslip, not every route through which the founder could benefit from a growing business.

The wager suited the company he had built. Craft Silicon’s campus offered flexible hours, meals, medical cover and recreational facilities. Deloitte named it Kenya’s best company to work for in 2013. Budhabhatti said employees held stock options and that the business helped some acquire cars and homes.

The Craft Silicon Foundation extended the proposition beyond employees. Formed in 2009 and led operationally by Priya Sodi, it provides free digital-skills training, with solar-powered mobile computer classrooms taking equipment and internet access into underserved communities. The foundation’s website says its work spans four countries.

The salary, campus and foundation reveal the part of Budhabhatti that understands the force of a concrete gesture. One shilling travels further in public memory than a compensation policy. A computer bus entering a neighbourhood is easier to see than a corporate commitment to inclusion.

More than 1,000 employees need policies that apply without a conversation with the founder. Executives need written compensation and ownership terms. Clients need service standards that do not rise or fall with Budhabhatti’s personal involvement.

The founder’s word helped create Craft Silicon’s culture. The company has reached the size at which the written rule must become stronger than the word.

Little finds the customer who stays.

Craft Silicon moved from software behind financial institutions to a product ordinary customers could see in July 2016. It developed Little Cab with Safaricom and launched it into a Nairobi ride-hailing market already learning the spending power of global platforms.

Little arrived with local features. Riders could pay through M-Pesa, cards, bank transfers or cash. Safaricom supplied free in-car Wi-Fi. Access through a short mobile code was planned for people without smartphones. Lady Bug allowed women to request female drivers. Little took a 15 per cent commission from drivers at launch, below the rates associated with larger competitors.

The early contest rewarded discounts, driver incentives and the ability to lose money while buying market share. Retail passengers could keep several applications on one phone and choose whichever offered the cheapest trip at that moment. Budhabhatti later described Little’s shift towards corporate transport as an act of survival.

A company buying employee transport behaves differently from a commuter chasing a promotion. It must authorise users, allocate departmental budgets, approve journeys, consolidate invoices and account for expenditure. Little built the administrative system around the ride.

By January 2025, Little told Rest of World that it had more than 5,000 corporate customers, two million users, over 200,000 registered drivers and annual gross merchandise value of roughly US$30 million. Gross merchandise value is the value of bookings passing through the platform, not Little’s revenue or profit.

Rest of World found that Little still trailed Uber and Bolt in overall daily rides in Nairobi. Budhabhatti said Little’s corporate business was larger than those competitors’ combined. Named customers described the attraction in practical terms. Aga Khan Hospital could set budgets and approvals by department, then use the reporting to monitor transport spending.

Little had stopped asking how to become the cheapest ride on every phone. It became part of the employer’s operating system.

That position grows more slowly. A corporate contract takes longer to win than a retail download, and local relationships do not automatically transfer across borders. Global competitors can improve their own administrative products. A corporate client can also demand more customisation, slower payment terms and service guarantees than an individual rider.

Little has widened the same platform into logistics, food, payments and entertainment. In April 2026, Budhabhatti said logistics generated about 30 per cent of its business, with ride-hailing contributing about 60 per cent. Logistics now covers motorcycles, three-wheelers, trucks and containers across Kenya, Uganda, Tanzania and Ethiopia.

The company delayed further expansion until existing markets could carry the marketing cost. That restraint reflects Craft Silicon’s history. Customer revenue built the parent company. Little could not assume that an investor would finance an indefinite discount war.

Its corporate niche is defensible because the product sits inside budgets and approval processes. It will remain defensible only if Little preserves service quality, driver supply and reporting advantages as competitors learn from it.

The founder steps aside.

Budhabhatti spent almost his entire adult working life in charge of Craft Silicon. In 2024, he handed day-to-day control of the African operation to new leadership while remaining Group Chief Executive and close to product development.

He later explained the decision with unusual candour. He had accepted projects without testing their strategic or financial value, allowed customers to influence prices too heavily, spread teams across work outside the company’s core competence and resisted partnerships that could have strengthened delivery.

“I am a typical nerd jerk,” he wrote. He was good at identifying what a customer might need and getting a product ready. Packaging, negotiation and saying no belonged to other people.

The new leadership introduced pricing and product matrices, narrowed project selection and opened more work to partners. Budhabhatti said Craft Silicon performed better after he left daily management.

The transition is sensible precisely because it does not pretend that the founder has become a different person. Craft Silicon kept him close to product ideas while moving pricing, delivery and operating discipline towards executives whose strengths lie there. His informal title, Head of Product Dreams and Designs, describes the bargain more accurately than a conventional corporate label.

Delegation will be measured by the decisions other managers can make against his preference. They must be able to reject a product he loves, price a contract above what he would accept, end a project he wants to rescue and promote a leader he did not personally choose.

Budhabhatti still carries group authority, ownership influence and public identity. Employees, clients and partners can therefore continue treating him as the final court of appeal even when an organisation chart assigns responsibility elsewhere.

The operating company becomes independent when a decision remains settled after the founder disagrees with it.

The promise reaches court.

One spoken commitment has already followed the company into a courtroom.

In October 2025, Kenya’s Employment and Labour Relations Court found that Craft Silicon and Little had unfairly dismissed Ronald Otieno Mahondo, Little’s founding general manager, and failed to honour a promised one per cent stake in Little.

Mahondo joined in 2016 to help establish the ride-hailing business. The court accepted electronic evidence of a meeting in which Budhabhatti acknowledged the equity award. It granted Mahondo KSh1.02 million in compensation for unlawful and unfair termination and US$750,000 for the stake. The dollar award used the US$75 million company value placed before the court and not substantively contested by the respondents in the proceedings.

Craft Silicon and Little challenged the judgment. On 5 June 2026, the court granted a stay of execution pending the intended appeal, conditional on KSh1.02 million being deposited in a joint interest-earning account. No reported appellate decision had displaced the 2025 judgment by 21 September 2026. Enforcement remains paused while the trial court’s findings await appellate review.

The appeal may change the legal outcome. An executive helping to build a new company should not have to rely on an audio recording to establish whether he owns part of it.

Start-ups use equity to recruit people they cannot pay at the level of established companies. A promise made across a table can feel sufficient when the founder and employee are working towards the same launch. The risk appears later, when the company becomes valuable, the relationship deteriorates or the parties remember the bargain differently.

An enforceable employee-ownership system needs a written grant, board approval, the number or percentage of shares, vesting conditions, treatment on departure, tax advice and an ownership register the recipient can inspect. A dismissal requires a valid reason and a fair process. Those controls protect the company as much as the employee.

Craft Silicon sells record-keeping systems to regulated institutions. Its internal commitments deserve the same discipline it builds into a bank’s accounts.

The next interface.

Budhabhatti has resisted turning Craft Silicon into a bank because ownership of a financial institution would place the company in competition with customers buying its technology. Little has instead become a laboratory beside the banks.

MySalary gives employees access to earned wages before payday through participating employers and financial institutions. Spotit supports purchases paid over time. TouristTap allows a visitor with an eligible contactless card and phone to pay a Kenyan merchant who receives the money through a local bank account or mobile-money channel.

TouristTap brings Budhabhatti back to a familiar type of break in the system. A tourist may carry a card accepted around the world, while a curio seller or small tour operator accepts only local payment rails. Craft Silicon developed the product with KCB and Visa, and Kenya’s tourism ministry endorsed it publicly in April 2026.

Its commercial test begins after the endorsement. Visitors must understand how to use it. Merchants must receive money reliably. Charges and exchange rates must be visible. Refunds and disputes must have a clear owner. Craft Silicon must show how many visitors and merchants transact repeatedly rather than counting downloads or launch partners.

Budhabhatti had been experimenting with machine learning for credit scoring before generative artificial intelligence became a mass product. He told Frontier Fintech that the early results appeared useful at the time but looked rudimentary beside what neural networks could later produce. Craft Silicon began treating artificial intelligence as part of the banking platform rather than a separate product attached to it.

His proposed first use was deliberately narrow. Before logging in, a customer would ask a bank’s assistant public questions about products or services through text or voice. After authentication, the same interface could retrieve personal information such as an account balance without requiring the customer to navigate a succession of menus. It would not initially make decisions, dispense financial advice or move money. Small transactions such as buying airtime would come later, after the customer had learnt when the system could be trusted.

A wrong answer about branch hours is an inconvenience. A wrong balance, repayment instruction or transfer can deprive a customer of money. Staged deployment gives the bank time to test retrieval, authentication, language performance, consent and dispute handling before conversation becomes execution.

Craft Silicon’s current SmallTalk product supports text and speech, multiple languages and installation within an institution’s own environment. The company markets Nimble as its newer API-first core-banking platform and dates its AI-based core-banking work to 2021. In the Frontier Fintech interview, Budhabhatti said Nimble’s conversational interface was being designed to guide staff through functions including account opening, card maintenance and reports.

A conversational layer connected to banking records cannot behave like a general chatbot. It must distinguish an answer retrieved from a verified account from a generated explanation. Permissions must restrict which balances, transactions and identity details the system can reach. The bank needs logs showing what the customer asked, which data the system used and what answer it supplied. A person must be able to correct an error before a wrong instruction causes financial loss.

Budhabhatti described the reach of Craft Silicon’s systems across more than 150 million end users as an advantage in developing artificial intelligence. Software reach is not ownership of customer data. A technology provider processes information under its contracts with the institution controlling it. Kenya’s data-protection rules require a lawful purpose, limit further use that is incompatible with the reason for collection and give customers rights over their personal information. A bank cannot turn ten years of transactions into training material merely because its software vendor can reach the database.

Banks’ reluctance to send customer records outside their data centres therefore shapes the product. SmallTalk offers on-premise deployment, and Nimble can run on-site, in the cloud or through a hybrid arrangement. Keeping data inside the institution does not by itself make the model safe. The bank still needs to know which records enter it, which model receives them, whether prompts are retained, who can retrieve the output and how inaccurate answers are challenged.

Budhabhatti expects an integration layer to assemble information from core banking, customer records and other systems before an artificial-intelligence service can use it. African banks used similar middleware to connect rigid core systems to mobile and USSD channels. That route can accelerate adoption, but each new layer also creates another dependency to secure, maintain and eventually replace.

His larger fear concerns Craft Silicon’s place in the value chain. The most advanced models, scarce computing capacity and highly paid specialists are concentrated among companies in the United States and China. An African software company can end up installing a foreign model for a bank, adding a margin and surrendering the creative work to its supplier.

Craft Silicon’s defence lies in the part of banking that a general model does not know on its own. The company understands local institutions, integrates with their records, supports systems in their markets and can test products through Little. Those relationships create access to problems, not an automatic right to customer data or a guarantee that the answer will outperform a global platform.

Artificial intelligence also raises the cost of informal founder-led product development. Security, data protection, model performance, pricing and responsibility for customer loss must survive review by people empowered to slow a release Budhabhatti wants to accelerate. The Head of Product Dreams and Designs needs executives who can turn down a dream before a bank turns it into a liability.

The first bank bought software from Budhabhatti because it solved an operating problem. Banks considering his AI products will apply the standard he helped establish. The interface may be new. The answer must still reconcile with the account beneath it.

The last system.

Budhabhatti published My Mother, Who Taught Me Business Without Knowing It in 2026. The book returns to the home in Jamnagar before programming, Nairobi and Craft Silicon.

His mother had not attended business school or run a company. He writes that she taught resourcefulness by making one shirt serve three children, customer psychology through the way she bought a glass of sugarcane juice and courage through ordinary decisions made under financial pressure. Author proceeds support education and digital-skills work through the foundation.

Resourcefulness built Craft Silicon when capital and institutional support were scarce. It is no longer the only standard. A group serving hundreds of financial institutions must make its performance, leadership authority and ownership commitments easier to verify. The founder’s ability to improvise must sit inside controls that tell employees and clients what happens when he is absent.

Craft Silicon has already answered a question that once hovered over African technology. Financial infrastructure can be designed in Nairobi, exported across emerging markets and supported for decades. Little has answered another. A local platform can survive global competitors by becoming indispensable to a valuable group of customers instead of trying to subsidise everyone.

Professional managers must be able to impose pricing discipline when the founder wants the contract. Equity promises should be provable without a recording. Client outcomes and group performance need measures stronger than company anecdotes. An AI assistant must inherit the reliability of core banking rather than the looseness of an experiment.

Budhabhatti returned to Nairobi because a bank program was unfinished. He completed it in a friend’s apartment and spent the next 26 years building new systems around the same instinct.

The last one is a company in which the promise is written, the manager can overrule him and the code keeps running after he leaves the room.