Elly here with your afternoon dispatch. Gutentag to Yoco’s new CEO. Katlego Maphai and team, thank you for your service as the Americans would say. We’ll get into Yoco’s c-suite changes this weekend, but for today:

Capitec deployed AI to 5,000 employees who now use it four times daily, blocked R673 million in fraud, embedded an agentic AI in credit processing, and grew IT spending 17.5% to R3 billion. While MTN CEO Ralph Mupita collected a record R99 million pay package, MTN Ghana completed the structural separation of its mobile money business into a standalone fintech on March 31. The same week, Moody's surrendered its South African regulatory licence, and Nigerian telcos kept failing to stop the spam call epidemic drowning subscribers. In short, African banking and fintech are splitting in two directions: the banks automating intelligently while growing headcount, and the telcos restructuring to unlock fintech valuations they can't capture under telecom regulation.

Today’s edition is packed to the rafters, so let’s dive right in.

Capitec saved R673 million in fraud prevention during their last financial year using AI-driven models. If their total IT expenses (excluding salaries) grew 12% to reach just over R3 billion, what percentage of their IT budget would the fraud savings represent?

A) 18.2%
B) 22.4%
C) 26.7%
D) 31.1%

(Answer at the bottom)

Why it matters: Capitec grew IT expenses 17.5% to R3 billion, deployed AI tools to 5,000 employees averaging four daily interactions, embedded an agentic AI in business credit processing, and prevented R673 million in fraud losses. Then the bank hired 711 more people, growing total headcount 4.3%. While the global narrative insists AI means job cuts, South Africa's largest retail bank is proving you can scale AI and employment simultaneously. That's not just a feel-good story. It's a deliberate capital allocation strategy that protects customer trust while automating the repetitive work that burns out good employees.

Zoom in: Cloud computing fees jumped 25% as Capitec scaled data capabilities. The bank's Pulse AI gives client support agents real-time contextualised information. AI-driven fraud models blocked 131,000 fraudulent beneficiaries and stopped 394,000 scam payments. Generative AI now runs in compliance operations. An agentic AI embedded in business banking handles credit processing with plans to scale across the division. CEO Graham Lee told shareholders AI "is not a future aspiration, it is already at work." The bank's 15.3 million app users processed 35% more digital transactions. Yet Capitec deliberately elevated model risk management to a tier-1 risk and flagged AI black boxes and agentic systems as emerging threats alongside cybersecurity and geopolitical volatility.

The Bigger Riskier Picture: Without knowing the larger context of Capitec’s total fraud and spam numbers - it’s difficult to say if their model is truly effective. Think about this: Blocking 394 000/500 000 scam payments is very different from 394 000/1 million or 5 million.

But, while the upside is clear, the approach introduces meaningful risk. Let’s look at this from a customer perspective, reliance on AI in areas like credit decisioning and fraud intervention raises concerns around incorrect decisions, bias, or even lack of explainability, particularly where outcomes materially affect financial wellbeing.

And then from a data perspective, systems like Pulse rely on analysing sensitive behavioural and transactional data in real time, which heightens the importance of strong governance, security, and clear consent frameworks. In the South African context, under South Africa’s Protection of Personal Information Act (POPIA), organisations must ensure personal data is processed lawfully, minimally, and securely, and cannot rely solely on automated decision-making where it has significant legal or financial consequences for customers.

This becomes even more critical as Capitec expands into agentic AI. Think of agentic AI like a smart digital assistant that doesn't just answer your questions, but can actually figure out a plan, use different tools, and finish a whole project for you without needing to be told every single step. Here systems can take actions with greater autonomy. The safer conclusion is that agentic AI in banking is acceptable only when the bank can prove rails are firmly in place: data protection, model governance, explainability, monitoring and human override. Ultimately, yes, Capitec appears to be leaning aggressively into AI-led scale, but long-term success will depend not just on capability, but on how transparently and responsibly these systems are designed, governed, and even trusted by customers.

Why it matters: MTN Ghana completed the structural separation of its mobile money business on March 31, creating MobileMoney Fintech Ltd as a standalone entity. The move follows Ghana's Payment Systems and Services Act requirement that telecom operators run financial services through separate legal entities. For MTN Group, this marks another step in a multi-year strategy to unlock fintech valuations that Mastercard's $5.2 billion 2024 investment validated but that remain hidden in telecom stock prices. The pattern is super clear: carve out mobile money operations country by country, create standalone entities with their own governance and reporting, and position them for separate investment or eventual listing.

State of play: MTN Ghana posted strong Q1 2026 results with service revenue up 35.7% to 7.3 billion cedis and profit after tax rising 46.8% to 2.5 billion cedis. Active MoMo users reached 18 million, up 4% year-on-year. MTN Nigeria's service revenue jumped 54.9% in constant currency, driven by a stronger naira and subscriber gains. But fuel costs are threatening to slash $102 million from profits as diesel and petrol inflation hits operations across 25,000 base stations. Meanwhile, CEO Ralph Mupita's pay package hit R99.3 million for FY2025, up 53% from R64.8 million, as long-term share incentives vested at R202.20 versus R124.60 in the prior cycle. The pay rise comes nearly a year after MTN's remuneration report failed to secure 75% shareholder support at the 2025 AGM, with 40.82% of votes cast against.

What’s next?: MTN's fintech separation is a template for how African telcos can extract value from mobile money operations that regulators increasingly want housed outside telecom structures. This weekend, we’ll get into IPO’s on the horizon - Opay and Airtel Money. Uganda and Ghana have completed the process. Nigeria's N152 billion spin-off proposal awaits final approvals.

The structural complexity is significant because each market has different regulatory requirements, ownership structures, and tax implications. But the strategic logic is consistent: mobile money generates hundreds of millions in revenue and processes massive transaction volumes, yet gets buried in telecom P&Ls where investors price it at telecom multiples instead of fintech valuations.

Separating the businesses lets MTN attract fintech-specific investors, build dedicated management teams, and eventually pursue separate funding or listings. For African banks watching telcos become payment giants, this separation makes MTN's fintech arms easier to partner with or compete against. You now know exactly where the mobile money operations sit, who runs them, and what their numbers look like.

Why it matters: Nigerian subscribers keep getting spam calls because telecom operators failed to implement adequate number portability protections and SIM registration enforcement. The problem got worse after number portability launched, making it easier for scammers to spoof legitimate numbers. Telcos have the technical capability to filter robocalls and implement caller verification. They just haven't prioritised it.

Catch up: Number portability was supposed to increase competition and give subscribers freedom to switch networks while keeping their numbers. What happened instead? Scammers exploited the system to register throwaway SIMs, port numbers to new networks, and overwhelm subscribers with robocalls selling everything from quick loans to investment scams. The spam got so bad that people stopped answering calls from unknown numbers, which defeats the purpose of having a phone.

So what: This isn't just about annoying calls. Spam erodes trust in the entire telecom infrastructure. When people stop answering their phones, legitimate businesses can't reach customers, banks can't verify transactions, and government agencies can't deliver services. For financial institutions relying on mobile channels for customer verification, the spam crisis creates a security problem: how do you distinguish a legitimate call from your bank versus a scammer spoofing the bank's number?

The fix requires telcos to implement proper STIR/SHAKEN caller verification protocols, regulators to audit network traffic instead of just collecting licence fees, and SIM registration databases that actually get checked before numbers activate. Until that happens, Nigerian fintech companies building on SMS and voice channels are building on infrastructure that subscribers don't trust.

And now for a breather while you mull on the news. If you’re using Claude, ChatGPT, Kimi or whatever LLM of your choice. Wispr Flow is a lifesaver. Imagine having a 24/7 intern, but instead of verbally giving them instructions, you have to type up every task you need them to do. It would take twice as long to get anything done. Speaking can instantly provides much more context than writing does. So, instead of typing up all the context you need, you can simply say it. That’s where Wispr Flow comes in. Turn your unstructured thoughts into structured instructions for any task.

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

Capitec's former CEO Gerrie Fourie is returning to the board less than a month after stepping down, overseeing the bank's international expansion strategy while new CEO Graham Lee runs day-to-day operations

Shiprazor raised R55 million to expand its logistics platform connecting South African retailers with courier networks for same-day delivery

Postbank is shifting services away from South African Post Office branches to new digital and retail banking channels as the postal service restructures

MTN Nigeria service revenue jumped 54.9% in constant currency on stronger naira and subscriber growth, while fuel costs threaten to slash $102 million from profits as diesel inflation hits 25,000 base stations

Breet launched a $10,000 equity-free grant for African builders ahead of Africa Technology Expo 2026.

YC-backed fintech Grey registered as a payment service provider in Canada, expanding its cross-border payment corridors beyond its core Nigeria-US route

Airtel Africa targets a $2 billion London IPO for its mobile money unit as African telcos race to unlock fintech valuations separate from connectivity businesses

Capitec Annual Report FY2025 - Deep dive into South Africa's largest retail bank's AI deployment strategy, fraud prevention results, and responsible AI framework

MTN Ghana Q1 2026 Results - Service revenue up 35.7%, mobile money separation completed, and fintech performance metrics across West Africa's third-largest operation

MTN Group Remuneration Report FY2025 - Executive compensation structure, long-term incentive vesting details, and shareholder voting patterns on pay packages

In the week of May 7, 1978: The first computerised cash dispenser linked to a central banking system went live in South Africa at Barclays Bank in Johannesburg. The machine, manufactured by De La Rue, could only dispense R10 and R20 notes and required customers to use special cheques that had to be ordered in advance. The system was so primitive that bank staff had to manually verify each transaction the following day by checking paper records. Within five years, Nedbank and First National Bank rolled out their own ATM networks, and by 1985 South African banks processed more automated transactions per capita than any other developing economy. The funny part? Customers initially refused to use the machines after hours because they didn't trust leaving the bank with their money in their pockets. Security guards had to be stationed at ATMs to convince people the cash was real.

Brain Teaser Answer: B) 22.4%

Capitec's IT expenses (excluding salaries) were just over R3 billion. The reported fraud savings of R673 million represents 22.4% of that total IT budget, showing that their AI investment in security alone nearly paid for a quarter of their entire technology infrastructure spend. Incredible.