Why Kenyan banks are ripping out the payment pipes behind their mobile apps.

On 28 October 1943, with the bombed chamber of the House of Commons still lying in ruins, Winston Churchill urged Members of Parliament to rebuild the room on its old, cramped oblong footprint, deliberately too small to seat every MP at once so that major debates would keep their crowded urgency. "We shape our buildings and afterwards our buildings shape us," he told the House.

Payment and core banking software shapes a financial institution in the same way. An architecture chosen fifteen years ago quietly dictates how many months a bank needs to launch a new checkout line, trace a missing transfer, or plug into a central bank switch. On 27 October 2026, I&M Bank Ltd announced it would rebuild its core payment infrastructure on an artificial-intelligence-native payment hub built by Otoma, a New York technology firm founded by the veterans behind Fundtech - now part of D+H , starting with its Kenyan operations. Jamie Loden, I&M’s chief operating officer, said the overhaul is meant to "simplify how we build, change and operate payment services".

For most of this year, banking conferences in Nairobi have focused on the front of the house, pitching commercial banks as lifestyle platforms and digital marketplaces that bundle third-party services. I&M’s contract points down into the basement. Kenyan banks spent the past ten years polishing the customer screen, and the next decade will turn on whether they replace the settlement machinery underneath.

Putting new mobile apps over old payment pipes.

Kenyan retail banking moved onto mobile phones faster than almost any analyst predicted. By 2021, Equity Bank Limited reported that 97% of its customer transactions took place outside physical branches, led by its mobile app. Because African banking markets often combine dozens of licensed lenders with sparse rural branch networks, Equity was able to multiply transaction volumes for years without pouring concrete for new branches. Rivals built large mobile credit books inside their own balance sheets through partnerships with Safaricom, including M-Shwari at NCBA Group and KCB Bank Group M-PESA at KCB, before NCBA spun out LOOP as a standalone digital subsidiary. In September 2026, Absa Group went a step further with Absa Next, a mobile app any Kenyan can open using a national ID card without holding a traditional Absa bank account.

Standalone challenger banks have struggled to displace incumbents in Kenya largely because established banking groups built their own neobanks first, backing new mobile apps with deposit-taking brands that savers already trust.

Across I&M’s 1.1 million customers in the first half of 2026, 92% were digitally active, channelling KSh 15.7 billion in financing through mobile and online tools. Digital lending and ecosystem partnerships generated 21.7% of the bank’s retail and business banking operating income, up from 14% a year earlier, helping lift group profit after tax by 22% to KSh 10.2 billion.

Even with that growth, I&M is replacing the payment hub underneath those channels to cut failed and duplicate transfers, track money in real time as it moves across rails, resolve disputes faster when a payment hangs, and shorten lead times when plugging into new payment networks. Over a fifth of I&M’s retail and business operating income now flows through digital journeys whose underlying payment switch the bank has publicly decided to rip out. Customers see a modern mobile app on the glass right up until a transfer fails on a Friday evening and the reversal disappears into a batch-era queue.

Hitting the ceiling of wrapper modernization.

For a decade, commercial banks modernized by wrapping their legacy cores. Engineers placed an API gateway or a smartphone interface on top of a payment engine built for overnight batch processing. While the customer screen looked fresh, the underlying data schema, release calendar, and failure points barely moved. In a 2026 technical paper, Otoma sets out the cost of that approach. Every new mobile money link, QR standard, or cross-border rail requires another custom translation layer until middleware patches become the bank’s working architecture. Changing a single routing rule triggers months of manual regression testing across bank and vendor teams. Legacy software vendors profit from keeping old hubs on life support through paid upgrade cycles and custom integration tickets. Once every minor tweak requires a vendor change request, product managers stop asking for small fixes and bend their commercial products around the limits of the software.

Otoma’s own leadership built much of that earlier generation of banking plumbing. Otoma’s chairman, Reuven Ben Menachem , founded Fundtech, while Otoma’s founder, Zack Yaniv, ran Fundtech’s global payments division. Hundreds of international banks bought Fundtech’s payment hubs before the company was rolled into Finastra in 2017, and much of that code still processes bank transfers today. The executives now selling AI-native replacement hubs built the exact generation of software they are asking banks to retire, arguing that "platforms once considered modern can become tomorrow’s legacy".

Regulatory shifts in Nairobi and abroad are forcing banks to confront that legacy code. Swift’s cross-border messaging network completed its mandatory ISO 20022 cutover in November 2025, and the detailed remittance data carried inside those messages cannot be used by a bank whose internal payment hub truncates the fields. At home, the Central Bank of Kenya and the National Treasury published a draft national payments policy and bill in September 2026 proposing a national instant payment switch, mandatory open API rules, regulator-enforced interoperability, and standalone licences for payment initiation and account information providers. Two months earlier, in July 2026, Kenya gazetted its virtual asset regulations, establishing a licensing regime for stablecoin issuers and introducing an on-chain settlement rail that legacy banking switches were never designed to handle. Real-time clearing leaves a bank milliseconds, rather than overnight batch windows, to spot fraud, manage treasury liquidity, and reverse a broken transfer. Bolting an instant payment rail onto a batch-era hub means the entire transaction still moves at the speed of the slowest pipe in the basement.

Turning the bank into a commercial convenor.

Over the next decade, a competitive commercial bank will operate much like a well-run shopping centre. The landlord provides the physical structure, security, and daily foot traffic so dozens of independent merchants can sell to the same shopper under one roof. In banking, the institution provides regulated payment rails, working capital, deposit custody, and verified KYC identity, while external partners plug in specialized services that let a customer book an airline ticket, buy travel cover, and split the cost over three months inside one checkout flow. Payments, credit, and deposits stop operating as siloed departments and work as continuous movements of value.

Autonomous software agents push that model further. As customers authorise AI agents to pay suppliers, renew policies, or sweep idle cash into money-market funds on their behalf, a bank’s job shifts from selling standalone financial products to executing a customer’s end-to-end instruction. Building a single configurable payment hub that runs across multiple countries also changes the unit economics of regional banking, allowing groups like I&M to amortise software builds across smaller East African subsidiaries in Rwanda, Tanzania, and Uganda.

A shopping centre landlord that takes three months to unlock a new shopfront or four weeks of regression testing to change a directory sign will lose tenants to faster rivals. When a software agent initiates a supplier payout, it requires an immediate cryptographic receipt confirming the funds settled, or a machine-readable error code explaining why the transaction stopped. In Otoma’s framing, the payment layer is where a bank either defends its primary relationship with the customer or surrenders the screen to an aggregator.

Rebuilding around instant service recovery.

Stripping out legacy payment pipes forces a bank to separate enduring financial principles from temporary local habits. Moving cash in and out through neighbourhood agents was M-PESA’s foundational principle, whereas the early "send money home" marketing campaign was a local Kenyan practice that failed when copied blindly into banked markets like South Africa. Transparent pricing, easy access, consumer protection, and fast service recovery govern financial trust in every market.

Of those principles, service recovery is the one old bank plumbing breaks every day. Nothing burns through decades of institutional brand equity faster than a failed transfer that debits a customer’s account without explanation, takes five working days to reverse, and leaves the customer bounced between the bank’s call centre, the payment switch, and the merchant’s cashier. When money moves in real time, fixing a broken payment has to happen in real time too.

Delivering that standard requires internal engineering teams schooled in data analytics, rapid third-party API integration, and vendor risk management who do not have to wait six months for an overseas supplier to write a patch. As shared payment switches turn basic money movement into a public utility, banks compete on what they build on top of the rail. Putting a configurable payment hub in the hands of a bank’s own technologists lets the institution share standard clearing infrastructure while tailoring checkout and credit workflows for its own corporate and retail clients.

Retraining staff counts as much as buying the software. Dropping AI tools into a technology department changes very little if compliance officers, product managers, and operations clerks keep following paper-era sign-off chains, making core modernization as much a job for a bank’s chief people officer as for its chief technology officer.

Managing the migration risk without a TSB outage.

Ripping out a live payment hub while millions of customers are transacting is one of the most dangerous operations a bank board can sign off on. In April 2018, when Britain’s TSB migrated customer accounts onto a new core platform over a single weekend, the system buckled immediately, locking up to 1.9 million customers out of branch, telephone, online, and mobile banking. By December 2022, UK financial regulators had fined TSB £48.65 million on top of £32.7 million the bank paid out in customer redress.

Avoiding a TSB-style outage comes down to how a board sequences the rollout. Otoma outlines three migration paths, allowing a bank to swap out the legacy hub in stages, stand up a single payment line alongside the old switch first, or use modular components to speed up an internal engineering build. Starting in Kenya gives I&M the chance to prove each payment rail in live production before switching off the old pipes or rolling the hub out to its regional subsidiaries.

Two governance risks still require close oversight from the board and the Central Bank of Kenya. First, escaping a legacy vendor’s fee treadmill achieves little if the bank locks itself into a new proprietary black box; a bank controls its own roadmap only if it retains portable ownership of its payment configurations, data mappings, and automated test libraries. Second, putting artificial intelligence inside live payment operations raises a direct supervisory question over who is legally accountable when an algorithm flags a compliance exception or reroutes a stalled wire. Otoma says its embedded AI agent, named Tom, drafts payment flows, writes test scripts, diagnoses failed transactions, and generates plain-language explanations of what broke. When a central bank examiner audits a failed payment run, the bank will be judged on that written explanation and human sign-off rather than on how quickly the code was generated.

Because I&M has publicly named the operational metrics it wants to move such as failed and duplicate transfers, payment visibility, dispute resolution speed, and time to market for new payment lines, publishing those numbers twelve months after go-live would give East Africa’s banking sector its first hard benchmark on what a rebuilt payment hub delivers.

Sitting at the same table to set the rules.

Kenyan banks won the past ten years by putting mobile apps in every customer’s pocket. Winning the next ten means fixing the settlement pipes behind the screen so a bank can convene outside partners and launch new workflows at the pace its customers demand. Building a glossy marketplace app on top of brittle batch-era plumbing leaves a bank with a fragile facade, while rebuilding the plumbing without opening the business to external partners reduces the bank to a low-margin pipe.

Even the cleanest internal payment hub cannot fix cross-bank friction on its own. National instant payment switches, open API mandates, and new payment-initiation licences only work when commercial banks, fintechs, mobile network operators, and clearing switches agree on shared connection rules, portable customer consent, and clear liability when a transfer stalls between two institutions. Those rules have to be hammered out around a shared industry table.

Churchill insisted on rebuilding a cramped Commons chamber because he knew the physical room would shape how future generations governed. By choosing which payment foundations to pour today, Kenya’s bankers are deciding what kind of financial institutions they will be able to run in 2035.