The hidden costs of agentic workflows are beginning to fracture traditional software business models. As the 'token amplification' effect creates a massive margin problem for vendors, we examine the inevitable shift away from seat-based SaaS pricing. Elsewhere in this edition, we parse why payments leaders are prioritizing active orchestration over simple PSP redundancy, and unpack the new three-year regulatory roadmap for South African fintechs.
The 'AI bill shock' we've been tracking—which recently drove a 22% drop in enterprise LLM token spending—has a mechanical root cause: the '100x problem.' While base model costs are falling, 'token amplification' within agentic workflows—where a single prompt triggers dozens of autonomous operations—is pushing SaaS vendors with traditional seat-based pricing into negative gross margins. This dynamic is now forcing a rapid shift toward usage- or outcome-based pricing models.
Why it matters
We've noted operators heavily scrutinizing AI ROI; this exposes the structural flaw driving that tension. For anyone building or buying AI-powered software, architectural decisions are now financial landmines. The SaaS pricing model as we know it is becoming untenable in the agentic era, requiring internal cost-aware technical strategies to prevent runaway expenses.
As we track the race by Visa and Nuvei to build 'Know Your Agent' (KYA) identity scoring for programmatic commerce, another structural flaw is emerging: dispute management. Current chargeback infrastructure relies on human-generated evidence like checkout clicks and receipts. Because autonomous purchases leave a completely different data trail, the industry is now pushing for a new 'evidence layer' to capture verifiable proof of machine intent and delegation within the transaction record.
Why it matters
This adds dispute architecture to the growing list of legacy systems that break when the customer is a machine. As the agentic commerce pilots we've covered scale into production, the lack of a purpose-built chargeback framework poses a massive risk to both merchants and acquiring banks. For infrastructure builders, designing this evidence protocol is the next major commercial opportunity.
A new PYMNTS Intelligence report surveying 60 US CFOs finds that 85% believe automation is crucial for improving payment speed and security. The study revealed a direct link between payment friction and revenue loss, with 55% of CFOs reporting that fraud or security controls had caused delays that negatively impacted customers. Firms with high friction estimated losing 1.92% of annual revenue to delays and errors, compared to just 0.31% for low-friction firms. AI/ML-powered real-time fraud scoring was identified as a critical solution.
Why it matters
This data provides hard numbers to back a strategic imperative: payments friction is a direct hit to the bottom line. For operators in merchant tech, this report is ammunition, quantifying the ROI of investing in automated, AI-driven fraud detection and straight-through processing. It shifts the conversation from a cost center (fraud prevention) to a revenue driver (friction reduction), which is a powerful framing for selling advanced payment infrastructure to enterprise clients.
Top-performing companies are not just adding more payment providers for simple failover, according to a new PYMNTS and Spreedly report. Instead, they are gaining a competitive edge by implementing intelligent orchestration with real-time monitoring, dynamic routing, and sophisticated rule updates. The study found that while 89% of firms have failover systems, only 47% achieve high transaction approval rates, revealing a significant gap between basic redundancy and true payment resilience.
Why it matters
This is a critical insight for any operator. The data shows that simply adding more PSPs to your stack doesn't guarantee better performance; it's the intelligence layer on top that drives revenue. For your work, this reinforces that the future of merchant tech lies in sophisticated, data-driven orchestration that can dynamically route transactions to maximize approval rates, reduce friction, and adapt to different market conditions—a key competitive advantage in both South Africa and broader African markets.
In a new interview, Robert Kraal, co-founder of Silverflow and former COO of Adyen, outlines his mission to rebuild payment processing with a modern, cloud-native platform. Silverflow provides a single API connection directly to card networks, aiming to eliminate the layers of legacy technology that add complexity and cost for acquirers and PSPs. Kraal emphasizes the importance of rich data and preparing the infrastructure for trends like agentic AI.
Why it matters
This is a direct challenge to the decades-old infrastructure that underpins most payment processing. By abstracting away the complexity of network connections and providing cleaner, richer data, platforms like Silverflow could fundamentally change the economics for payment players. It's a move to commoditize the lower-level infrastructure, allowing PSPs and acquirers to focus on value-added services—a trend that will eventually impact the South African market.
Visa is intensifying its push for 'Click to Pay' in 2026, integrating the express checkout solution directly with major e-commerce platforms like Shopify and BigCommerce. The initiative aims to streamline guest checkout by creating a network-native standard that allows shoppers to authenticate once and use their saved card details across any participating merchant, boosting conversion rates by reducing friction.
Why it matters
This is the card networks' answer to Apple Pay and other device-native wallets that have dominated the express checkout space. By embedding a standardized, tokenized solution at the platform level, Visa is attempting to own the guest checkout experience and prevent disintermediation. For merchants, it offers another tool to combat cart abandonment, and for the payments ecosystem, it's a key battle in the war for checkout dominance that will likely influence future standards in South Africa.
South Africa's Financial Sector Conduct Authority (FSCA) has published its regulatory plan for April 2026 to March 2029, providing the clearest timeline yet for major changes affecting the fintech sector. The roadmap details forthcoming rules on open finance, data storage and cloud outsourcing, licensing for payment services, and the transition to the new Conduct of Financial Institutions (COFI) Bill. While many rules are still in draft, the calendar gives firms a multi-year view of the compliance landscape.
Why it matters
This isn't just another regulatory update; it's a strategic roadmap for the entire South African fintech ecosystem. For operators like Yoco, it dictates the future operating environment for data management, third-party vendor governance, and the eventual shape of open finance. Understanding these timelines is essential for long-term strategic planning, resource allocation for compliance, and identifying competitive opportunities or risks stemming from the new regulatory architecture.
GoTyme Bank has become the only major South African bank to offer free PayShap transfers for all transaction values. This move positions the digital bank aggressively against competitors like Capitec, FNB, and Standard Bank, which all charge varying fees for the instant payment service. GoTyme's CEO stated that free, instant digital payments should be a standard feature of modern banking.
Why it matters
This is a direct price-based assault on the revenue models of incumbent banks and a powerful move to drive customer acquisition. By making PayShap transfers free, GoTyme is commoditizing a service others monetize, potentially forcing competitors to re-evaluate their own fee structures. For the SA payments landscape, this accelerates the push towards frictionless, low-cost digital payments and puts pressure on all players to justify their pricing.
The Western Cape High Court on Monday ruled that contracts awarded by the National Student Financial Aid Scheme (NSFAS) to four fintech companies for distributing student allowances were unlawful and invalid. The court cited procurement irregularities by NSFAS but found no evidence of corruption by the fintech providers themselves, who may be able to seek fair compensation for services rendered. The ruling has significant implications for how financial aid is disbursed.
Why it matters
This ruling underscores the high operational and legal risks of public-sector contracts in South Africa. For the fintech sector, it's a stark reminder of the importance of scrutinizing procurement processes. While the invalidation creates uncertainty, the court's distinction—blaming NSFAS's process, not the fintechs' conduct—and opening a path for compensation, sets an important precedent for providers caught in flawed government tenders.
Shares in trading platform Plus500 dropped 11% on Monday after its H1 2026 update revealed significant margin pressure. Despite a 12% rise in revenue, EBITDA grew by only 1%, with margins narrowing to 41% from 45% a year earlier. The company cited higher customer acquisition costs (CAC) and investments in its US expansion for the contracting profitability, and revised its full-year guidance from 'ahead' of consensus to merely 'in line.'
Why it matters
This is a textbook case study in the challenges of balancing growth and profitability in fintech. Plus500's experience shows how quickly rising customer acquisition costs can erode margins, even when top-line revenue is growing. It's a cautionary tale for operators and investors, underscoring the market's increasing scrutiny of unit economics and the shift away from a 'growth at all costs' mindset.
Shoprite Group announced on Sunday that its AI shopping assistant, Pixie, has seen 98% adoption among Checkers Sixty60's Xtra Savings Plus members within three months of its April launch. The company reports that Pixie, which uses loyalty data to predict shopping needs, has led to higher average order values and significantly faster shopping times, with one user completing a R1,500 order in 15 seconds. The news comes as competitor Pick n Pay also rolls out its own AI assistant, Penny.
Why it matters
This demonstrates a real-world, at-scale deployment of AI in South African retail, moving beyond pilots to widespread consumer use. The high adoption rate is a strong signal that consumers are ready to embrace AI tools that offer tangible convenience. For the local retail and tech landscape, this escalates the arms race for AI-powered commerce, putting pressure on all retailers to develop similar capabilities to compete on user experience and efficiency.
Wamkele Mene, the Secretary-General of the AfCFTA, stated that Africa loses $5 billion annually due to the costs and barriers associated with currency fragmentation and cross-border payments. The statement underscores that despite progress on trade agreements, the lack of harmonized payment systems remains a major structural obstacle to deeper economic integration on the continent.
Why it matters
This figure puts a hard cost on a problem that pan-African payment operators are trying to solve. It highlights the massive market opportunity for platforms like PAPSS and private-sector fintechs that can streamline cross-border settlement in local currencies. For anyone operating in African commerce, this is a reminder that the biggest hurdles—and opportunities—are often in the unglamorous-but-essential infrastructure layers of payments and logistics.
The '100x Problem' and the Crisis of SaaS Pricing Models AI agents are creating an 'efficiency penalty' for SaaS vendors. A single user request can trigger a cascade of operations, leading to 'token amplification' that blows up costs and makes traditional seat-based pricing unsustainable. This is forcing a rapid, painful pivot toward usage- and outcome-based pricing models.
Payments Orchestration Shifts from Redundancy to Intelligence Top-performing merchants are no longer just adding more payment providers for failover. Instead, they are winning by implementing sophisticated orchestration layers that use real-time observation, intelligent routing, and dynamic rules to maximize approvals and revenue. It's a shift from static architecture to dynamic, operational revenue management.
Pragmatic, Problem-Solving AI Defines African Adoption Across the continent, from Nigeria to South Africa, the focus on AI is practical. The emphasis is on solving local problems like financial inclusion, SME credit gaps, and trade friction, rather than chasing hype. This is driving the development of localized models, data strategies, and applications tailored to African market realities.
Agentic Commerce Demands a New Security & Evidence Layer As AI agents begin to transact autonomously, the industry is confronting a massive security gap. Traditional fraud and dispute management systems are breaking down, highlighting the urgent need for new controls around agent identity, verifiable intent, and a dedicated 'evidence layer' to manage disputes in a machine-to-machine economy.
South Africa's Regulatory Roadmap Comes into Focus The FSCA is providing clearer timelines for major regulatory shifts, including the COFI Bill, open finance frameworks, and new data rules. For fintech operators, this provides a critical window to prepare for significant changes to the compliance and operating environment over the next three years.
What to Expect
July 14, 2026—Major US banks, including Citigroup, Bank of America, and Wells Fargo, are scheduled to report their corporate earnings.
2026-2029—South Africa's FSCA will roll out new regulations covering open finance, data storage, and the COFI Bill.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
421
📖
Read in full
Every article opened, read, and evaluated
186
⭐
Published today
Ranked by importance and verified across sources
12
— 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