The bill for enterprise AI experiments is coming due. New data shows companies are burning an average of $11.5 million annually on AI with only 7% tracking the actual returns, prompting a massive CFO-driven crackdown on hidden inference costs and 'telemetry taxes'. Today's briefing unpacks those specific unit economics, tracks Microsoft's move to standardize agentic commerce, and looks at the strategic pivot by African digital lenders away from unsecured consumer credit.
Following yesterday's signals that enterprises are shifting from experimental AI to a cost-conscious phase, new data quantifies the accountability gap driving the CFO crackdown. A KPMG survey notes that only 7% of organizations have a formal process for establishing AI ROI. Separate analysis finds companies are spending an average of $11.5 million annually on AI, often without demonstrable returns, accelerating a reckoning over AI budgets and token costs.
Why it matters
This marks a crucial maturation point for the enterprise AI market. The era of speculative, 'magic bean' spending is ending, replaced by a CFO-led demand for hard, measurable returns. This fundamentally changes the vendor landscape, favoring solutions that solve narrow, specific problems with clear cost-to-value ratios over ambitious, open-ended platforms. For operators, it validates a disciplined, ROI-focused approach to AI adoption.
Nigerian-founded fintech Stabyl has emerged from stealth with $2.7 million in pre-seed funding, led by e-commerce giant Konga, to build institutional-grade foreign exchange (FX) infrastructure for Africa. The platform will use a central limit order book to connect banks, payment service providers, and other financial institutions, aiming to create a more efficient and liquid marketplace. Stabyl will support both traditional banking rails and stablecoin-based settlements to reduce friction in cross-border transactions.
Why it matters
Stabyl is tackling one of the most fundamental challenges in pan-African commerce: fragmented and inefficient FX liquidity. By creating a centralized infrastructure layer, it aims to reduce the high costs and long settlement times that plague cross-border payments. This could become a foundational 'plumbing' service for the entire African fintech ecosystem, enabling smoother trade and payments for everyone operating on the continent.
Adding to the wave of Nigerian fintechs securing upgraded banking licenses, TechCabal reports that payment giants Paystack and Flutterwave are strategically acquiring Microfinance Bank (MFB) licenses to evolve beyond simple payment processing. This shift allows them to accept customer deposits and offer credit, fundamentally changing their business model from being reliant on transaction fees to a more sustainable, interest-income model based on lending.
Why it matters
This trend signals the next stage of maturation for Africa's most developed fintech market. It's a strategic necessity driven by fee compression in the payments space and a desire to own more of the value chain. This evolution blurs the lines between fintech and traditional banking, accelerating the trend we've seen of Nigerian payment operators deepening their structural footprints, and will likely force further regulatory responses from the Central Bank of Nigeria that will set precedents for other African markets.
A new Connecting Africa report spotlights four fintech startups—Omnisient (South Africa), Nkwa (Cameroon), Hub2 (Côte d'Ivoire), and Zemi (Kenya)—as exemplars of the next wave of African fintech. These companies are distinguished by a focus on sustainable unit economics, disciplined product execution, and regulatory savvy. They are addressing deep infrastructure needs like SME working capital, data-led underwriting, and payroll infrastructure rather than focusing on high-burn consumer growth.
Why it matters
These companies represent a market shift from the 'blitz-scaling' narrative toward building resilient, profitable businesses that solve fundamental economic problems. Their models, which emphasize deep local insight and sound business economics, provide a valuable playbook for building successful fintechs in complex African markets, showing a clear path to long-term value creation beyond vanity metrics.
The era of mass, unsecured consumer lending by African digital lenders is ending. According to a TechCabal analysis, fintechs like Tala (which is restructuring) and Branch (which cut staff despite profits) are pivoting away from high-risk, standalone loan apps. The new focus is on embedded finance, lending to merchants with visible cash flow, and partnerships. The success of players like 4G Capital in the merchant lending space exemplifies this strategic shift toward more sustainable, data-driven models.
Why it matters
This represents a significant market correction. The 'growth-at-all-costs' model of acquiring millions of unvetted borrowers has proven unsustainable. The pivot to embedded and merchant-focused lending signals a maturation of the African fintech credit market, where profitability and sound risk management are now prized over raw user numbers. This creates new opportunities for B2B fintechs that can provide the data and infrastructure for this more secured lending model.
Despite being a global leader in mobile money adoption, Kenya's underlying payment infrastructure remains fragmented and inefficient. A new analysis argues that the country's next challenge is to build an integrated, instant, and low-cost system to connect its many siloed providers. The piece highlights the strategic importance of national switches like Kenswitch and Pesalink in creating the interoperability needed for the next phase of digital commerce.
Why it matters
This highlights a critical 'last mile' problem in a mature mobile money market. While consumer adoption is high, the back-end fragmentation creates friction and limits the potential for more complex digital services. For operators, this signals a significant opportunity in providing the infrastructure—the switches, settlement layers, and orchestration—that will unlock the next wave of growth and efficiency in East Africa's most advanced digital economy.
Major European fintechs are making significant moves into Africa, signaling a recognition of the continent's profitable payment corridors. UK challenger bank Monzo has launched direct in-app transfers to Nigeria in Naira. Meanwhile, Revolut is preparing for a full South Africa launch, with a waitlist that has already surpassed 100,000 registrations ahead of its planned 2028 market entry.
Why it matters
The entry of top-tier global fintechs like Monzo and Revolut validates Africa's remittance and payment corridors as first-class, high-potential markets, not just secondary expansion targets. This will inject significant competition for local and regional players, likely driving down costs for consumers and forcing incumbents to accelerate innovation. For operators, it's a clear signal that the African payments landscape is globalizing.
The Model Context Protocol (MCP) we've seen adopted for AI payment tools like MoonPay's virtual cards is now moving into enterprise retail. Microsoft has released its Dynamics 365 Commerce MCP server into public preview, creating a universal language for AI agents to interact with retail business logic. It allows commerce platforms to expose their capabilities—like product discovery, inventory, pricing, and checkout—once, enabling any compliant AI agent to access them in real-time without custom integrations. Early adopters like jeweler Michael Hill are already using it for AI-driven shopping experiences.
Why it matters
This is a significant move toward solving the 'N-to-N' integration problem in agentic commerce. Instead of building bespoke connections for every AI and every retail platform, MCP aims to create a standardized 'socket' for commerce. For the industry, it's a direct competitor to approaches we've tracked from Shopify (UCP) and others, signaling a race to define the standards that will govern machine-led retail. For operators, this could dramatically lower the friction of deploying AI sales and support agents across different platforms.
A new project called Warp Point, launched by tech journalists Wes Fenlon and Matt Sayer, aims to recreate the experience of the 'old internet'. The initiative is a conscious move away from algorithms, AI, and engagement-driven platforms, focusing instead on human curation, direct discovery, and fostering communities through older technologies like RSS feeds.
Why it matters
This project taps into a growing 'tech nostalgia' and a sentiment that something valuable was lost in the transition to the algorithmically-mediated web. While a niche movement, it's a useful counterpoint to the relentless push for AI-everything, exploring alternative models for content and community that prioritize human connection and serendipitous discovery over automated optimization.
Building on the shift toward AI agent unit economics we've been tracking, a new analysis details the hidden cost multipliers that inflate AI bills far beyond simple token pricing. Inference costs are driven up by the 'stateless agent tax' (re-feeding context for every task), the 'reasoning trap' (long, expensive thought chains), and the 'telemetry tax', which can add 40-200% to the inference bill for observability and logging. The author argues that measuring cost-per-successful-task is the only way to manage AI economics effectively.
Why it matters
This analysis provides a critical framework for any operator deploying AI agents, moving the conversation from capability to operational viability. For fintechs building AI-driven merchant services, underestimating these hidden costs can destroy margins and make products financially unsustainable at scale. This is essential reading for anyone budgeting for or building agentic systems, as it forces a more realistic and disciplined approach to architecture and vendor selection.
South Africa's Takealot Group, owned by Naspers, achieved profitability for the first time in fiscal year 2026, reaching $1 billion in revenue. The landmark year was driven by strong growth in its core e-commerce platform Takealot.com, its food delivery service Mr D, and the expansion of its logistics arm, Takealot Fulfilment Solutions.
Why it matters
Takealot's profitability is a major milestone for South African e-commerce, signaling that the market has reached a new level of maturity and scale. It proves that a large-scale, digitally-native retail and logistics operation can be profitable in the local context, setting a powerful benchmark for other players and likely intensifying the digital transformation race among traditional brick-and-mortar retailers.
Kenya's Electronic Tax Register (ETR) system has gone through three generations, evolving from standalone hardware in 2005 to the current software-based eTIMS system. A guide from POS provider Veira explains that this latest shift allows businesses to issue tax-compliant invoices directly from mobile phones or POS applications, transmitting sales data to the Kenya Revenue Authority (KRA) in real-time without needing a dedicated hardware box.
Why it matters
This technological evolution in tax compliance is a game-changer for merchant operations in Kenya. The move to software-based eTIMS dramatically lowers the cost and complexity for SMEs to become compliant, creating a major opportunity for POS and payment providers to embed this functionality directly into their offerings. It's a key example of how regulatory tech is shaping the merchant acquiring landscape in African markets.
AI Investments Face a CFO Reckoning Following massive spending on AI, a wave of reports shows CFOs are now demanding measurable ROI. Companies that can demonstrate value in narrow, specific workflows are thriving, while those with vague, experimental deployments are facing budget cuts. The era of 'AI for AI's sake' is over.
The 'How,' Not Just the 'What,' of AI Pricing Vendors are realizing that bundling complex AI features into flat-rate SaaS tiers is a recipe for margin compression. The sustainable model emerging is outcome-based or consumption-based pricing, where costs align directly with the value delivered, a shift procurement teams are now actively tracking.
African Fintech Infrastructure Matures and Attracts Global Players From FX liquidity platforms like Stabyl to major UK fintechs like Monzo and Revolut entering African corridors, the continent's payment infrastructure is moving into a new phase. The focus is shifting to building foundational layers for interoperability, efficiency, and cross-border settlement, attracting serious global competition.
Agentic Commerce Moves Toward Standardization As AI agents begin to transact, the need for common standards is becoming acute. Microsoft's new Model Context Protocol (MCP) for Dynamics 365 is a major step, aiming to create a universal language for agents to interact with retail systems, moving the market from bespoke integrations to a more interoperable future.
African Digital Lending Pivots to Embedded and Secured Models The era of mass, unsecured digital lending in Africa is winding down. Faced with high default rates, leading fintechs are restructuring and shifting their focus to secured lending, merchant cash advances, and embedded finance models that leverage visible income streams for more sustainable risk management.
What to Expect
2026-06-30—Stripe Tour Berlin 2026, focusing on AI-driven payments and orchestration.
2026-07-01—California's new allergen disclosure law (SB 68) takes effect, impacting restaurant POS and menu systems.
2026-07-31—Nigeria's 3rd Business Journal Fintech & Financial Inclusion Roundtable.
— The Merchant Desk
🎙 Listen as a podcast
Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.
Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste