Today on The Merchant Desk: the agentic commerce stack stopped being a thought experiment. Google's Universal Cart is live, Visa's running an Agentic Ready program through CEMEA banks, and Primer raised $100m to position AI as the operating layer for payments orchestration. Underneath it: Lesaka's first profitable year leaning on a stretched SA consumer, Peach drags BNPL into physical checkout, and Cape Town's minibus taxis go cashless on 1 June.
Yesterday's Kapron framing — that agents collapse merchant choice to the lowest-cost rail — now has a shipping product behind it. Google I/O 2026 (19 May) rolled out Universal Cart across Search and Gemini this summer, then YouTube and Gmail, monitoring prices and executing checkout autonomously via Agent Payments Protocol (AP2). The Universal Commerce Protocol expanded to Canada, Australia and the UK and into hotels (Booking, Expedia, Hilton, Marriott) and food delivery (DoorDash, Uber Eats, Square, Toast). The UCP Tech Council now includes Amazon, Meta, Microsoft, Salesforce and Stripe. Gemini 3.5 Flash — 4x faster than frontier models — is the inference layer, with Antigravity 2.0 as the agent IDE.
Why it matters
The vertical expansion matters more than the geographic one: hotels and food delivery are exactly the categories where SA merchants compete on margin and where agentic discovery will reroute demand. Two practical implications for any merchant tech operator: (1) catalogue cleanliness, real-time inventory APIs and structured product data become survival infrastructure, not nice-to-haves; (2) you have to decide whether to support UCP (Google), AP2, OpenAI/Stripe's ACP, and x402 — or be invisible. Dual-protocol merchants are already reporting up to 40% more agent-driven traffic. The networks (Visa, Mastercard, PayPal) want to be the trust layer; Google wants to own the cart. Both can be true, and merchants pay for both.
Lesaka Technologies posted Q3 2026 net revenue up 16% YoY and adjusted earnings of 180c per share, driven by an 81% jump in consumer segment EBITDA. Management raised full-year guidance to 550c–600c and expects the first profitable year since the 2022 rebrand. But two yellow flags: merchant division revenue slipped sequentially, and lending originations fell 22% YoY as Lesaka tightened credit standards. The Bank Zero acquisition close is the swing factor for FY guidance.
Why it matters
Lesaka has been the closest thing SA has to a Nu-style integrated consumer fintech narrative, and the consumer-segment unit economics are clearly working. But the divergence inside the business is exactly what to watch: scaling cheap-to-acquire grant-beneficiary consumers while the merchant book — Lesaka's structural moat against banks — softens. The 22% drop in lending originations alongside the fuel shock and household pressure suggests management is reading the macro the same way SA economists are. The real test arrives over the next two quarters: do loan losses spike, does the merchant ARPU recover, and does Bank Zero deposit funding actually arrive in time to underwrite continued growth? For anyone watching the SA acquiring league table, this is the cleanest read-through to how stretched the low-income consumer really is.
Peach Payments integrated PayJustNow's Pay-in-3 and Pay-in-12 directly into its POS device stack — no QR codes, no manual entry, native reconciliation. Globally the same week: Worldline and Klarna agreed to push BNPL across Worldline's online and in-store estate; Klarna launched a ChatGPT shopping search app; Affirm extended BNPL into Royal Caribbean travel bookings; US credit unions integrated BNPL natively after PYMNTS data showed 70% of Gen Z would use BNPL from their primary bank.
Why it matters
BNPL is no longer a button on a checkout page — it's becoming acquirer-level infrastructure. Peach's move turns a payment method into a merchant feature embedded at the terminal, which is exactly where Yoco, iKhokha, Ozow and the bank-owned acquirers now have to respond. Two structural reads: (1) the locus of BNPL competition is shifting from consumer brand (Klarna, PayJustNow, Mobicred) to who owns the merchant-side rails and reconciliation; (2) BNPL economics inside an acquirer stack are very different from standalone — the acquirer can subsidise the merchant fee against payment volume and use BNPL as a stickiness moat rather than a P&L line. For SA mid-market merchants, the question to ask vendors stops being 'do you accept BNPL' and starts being 'is it native and reconciled, or is it a referral?'
Three SA digitisation deadlines landed on the same desk this week. Codeta confirmed all Cape Town minibus taxis go cashless on 1 June 2026, with tap cards or the SAPAY app replacing cash across every vehicle (payment scanners and cameras included). Postbank set 31 August as the hard cutoff for SASSA Gold-to-Black card migration. SARS rolls out its Traveller Management System for foreign-registered vehicles on 1 June. Underneath all of it: a R302bn Social Development budget funding SASSA's biometric rollout across 432 offices and R13.8bn for Home Affairs to push Smart ID issuance from 167 to 750 bank branches.
Why it matters
This is the closest thing SA has to a forced-march digitalisation of its informal and grant economies, and the calendar matters. Three things to watch: (1) the wallet question — Cape Town taxis need a default wallet on day one, and whoever wins the SAPAY-equivalent rollout in other metros gets distribution to 15 million daily commuters; (2) the SASSA Gold migration creates a one-off window where 8m+ beneficiaries become acquirable customers as their card relationship is reset — Lesaka, MyMoolah, TymeBank and the big four all have a play; (3) the Home Affairs ID-at-bank-branch expansion turns retail banks into the de facto identity layer, which has real implications for fintech KYC costs and account-opening friction. The risk vector is cybersecurity — every downstream service inherits whatever architecture Home Affairs locks in.
Primer — founded by former Braintree/PayPal operators — closed a $100m Series C led by Sofina (total funding $170m), targeting US revenue growth from one-fifth to one-third of the book by 2028. The capital is going into Primer Companion, evolving from advisory tool to an autonomous agent that runs payment-routing experiments and approval-rate optimisation within merchant-defined guardrails. The platform captures 400+ data points per transaction across fragmented processor networks and handles 95% of customer payment volume on average.
Why it matters
This is the orchestration thesis crystallising into a fundable, autonomous-execution product. The economic logic is clean: in a multi-acquirer, multi-method world, 2–7% approval lift (the same band dLocal cited yesterday for network tokenisation) compounds into material revenue for any high-volume merchant. The defensibility play here is the unglamorous one — organisational memory. Transaction context, merchant-specific rules, historical retry patterns and audit trails are the moat; the model itself is interchangeable. For SA's enterprise merchants (retail groups, telcos, hospitality chains) running multiple acquirers across multiple BIN ranges, this is the architecture they'll be sold within 18 months. Expect Adyen, Stripe and the larger SA acquirers to either build, buy or partner here.
Four shipping announcements: Fireblocks launched an Agentic Payments Suite with built-in compliance and spend governance and joined the Linux Foundation's x402 group. AllUnity (DWS/Flow Traders/Galaxy JV) prepped SEKAU — the first Swedish krona stablecoin under MiCAR — and launched its own x402-based agentic payments product. MoneyGram became Tempo's anchor remittance validator with Stripe wiring stablecoin settlement into the flow. PayPal expanded PYUSD into 70 markets including African corridors with settlement compressed from days to minutes. Deloitte's forecast: stablecoins could carry 2.5% of US noncash and $200B+ in retail payments by 2030.
Why it matters
Two patterns to register together. First: stablecoin payments have decisively moved from crypto experimentation to mainstream rails, and the carriers are regulated incumbents (PayPal, MoneyGram, DWS) — not crypto-native players. Second: stablecoins and agentic payments are converging on the same infrastructure stack (x402, AP2, programmable wallets) because both need sub-second settlement and policy-enforced spend controls that card rails struggle to deliver. For SA and African operators, the PYUSD expansion is the most immediate signal — settlement compression on remittance corridors directly competes with the FX margins that fund a chunk of African fintech P&Ls. The harder strategic question: do you become a PYUSD/USDC merchant of record, build your own ZAR-backed equivalent, or partner with whichever issuer wins in the SARB framework? The first-mover window here is 12–18 months, not three years.
Nu Holdings posted Q1 2026 record numbers — $5B revenue, $871M net income (+80% FX-neutral), 135m customers, 29% ROE, 17.6% efficiency ratio. AI-driven engineering throughput is up 50% YoY with testing cycles cut 90%. And yet the stock dropped 4–6% on a small revenue miss against consensus and disclosed early-stage delinquencies. Same week: PagSeguro's credit book grew 36% YoY while deposit funding only grew 23% as SELIC cuts into margins; StoneCo flagged 2025-cohort churn from bundling complexity and 21.9% cost-of-risk; Paysafe's EBITDA margin compressed 130bps despite 10% revenue growth; Sea and MercadoLibre kept deliberately compressing fintech margins for ecosystem moves. Mercury — four years GAAP-profitable, conditional OCC charter — raised $200m at a $5.2B mark.
Why it matters
The market is now triangulating fintech valuation on capital efficiency, not customer-count or top-line growth. Nu's elite efficiency ratio doesn't immunise it from a wobble when provisions rise; PagSeguro's loan-to-fund mismatch is structurally riskier than its growth number suggests; StoneCo is teaching everyone that bundling done badly costs you a churn cycle. Two operator takeaways: (1) credit-book expansion without commensurate deposit growth is the trap to avoid — it's the variable that broke a string of LATAM fintechs in 2024–25; (2) Mercury's premium comes from owning the rails (charter), not feature velocity. For SA — where Lesaka is making the same credit-book-against-tightening-consumer bet on a smaller scale, and TymeBank, Discovery and Capitec are all competing on efficiency ratio — these LATAM tape-reads are the closest comparable benchmarks we have.
President Trump signed an executive order on 19 May directing the Federal Reserve and other regulators to review rules restricting fintech innovation, specifically including direct fintech access to Fed master accounts (Fedwire) and deposit services. A 120-day comprehensive review and 180-day recommendation deadline applies. Kraken received a master account in March 2026; Ripple, Anchorage and Wise have applications pending.
Why it matters
Master account access changes fintech unit economics structurally — direct Fed access removes the sponsor-bank layer, retains revenue on deposits and lending, and eliminates a fee tier that compresses MDR for any US-touching fintech. The second-order effects matter even for SA operators: (1) US fintech margin expansion enables aggressive cross-border product investment, which is competitive pressure on African remittance corridors; (2) precedent on non-bank access to central bank rails will inform the SARB's 'Beyond PayShap' work and the SADC-RTGS evolution; (3) if Wise lands a master account, the FX margin in SA-USD-SA corridors compresses further. This is one to track on the 17 September 120-day deadline.
Three datapoints triangulate restaurant AI moving from pilot to operations. Arcos Dorados (McDonald's largest franchisee, 20+ LATAM markets) reported Q1 2026 system-wide comparable sales +16%, with digital channels now 64% of system-wide sales and 30m+ loyalty members driving 25% of total sales. Sapaad Signals launched a 'pre-loss' intelligence system that monitors 18 EBITDA-critical KPIs in six-second cycles — early deployments report 11% revenue recovery from real-time intervention on voids, discount misuse and stockouts. The Saudi Restaurant and Cafes Owner Association signed Tabsense to roll an AI-native POS across 6,000 outlets. Solink also launched camera-based AI drive-thru intelligence to replace loop sensors.
Why it matters
Restaurant AI is now solving the problem POS systems never could: shifting loss management from end-of-day reporting to in-service intervention. The 11% revenue recovery figure (if it holds up across more deployments) is the kind of unit-economics number that breaks the typical pilot-to-production stall — chains can fund AI deployment from recovered margin within a single quarter. For SA hospitality groups (Famous Brands, Spur, Tashas, Yum Brands' local franchisees) and any merchant tech vendor selling into them, the competitive bar has moved from 'integrated reporting' to 'agentic operations.' Arcos Dorados is also the cleanest read on what loyalty-plus-digital can do at scale: 25% of sales attributable to loyalty members is the benchmark Pick n Pay's Smart Shopper, Clicks Clubcard and Discovery Vitality should be measured against.
Two M&A moves on the same day pulling in the same direction. NMI acquired Dwolla — its sixth acquisition in recent years — folding A2A, FedNow and real-time payouts into its white-label embedded stack ($700bn combined annual volume, Dwolla's CEO becomes NMI's COO). Mollie agreed to acquire GoCardless to combine cards, local methods and bank payments across 350,000 businesses on the back of 29% revenue growth to €147m in 2025; Mollie has also pushed €250m in Mollie Capital lending and launched business accounts. Same week, Worldline×Klarna and Adyen×Starling continued the trend.
Why it matters
The boundary between 'acquirer', 'orchestrator' and 'embedded banking provider' is gone. Any platform serving merchants in 2027 is expected to carry card, A2A, real-time, BNPL, lending and a working capital product on a single API. Two reads: (1) for mid-market SA acquirers and ISVs, the build-vs-buy clock is ticking — every quarter you don't have native A2A and real-time payouts, your enterprise pitch loses ground; (2) the M&A pricing reveals what the market values: Dwolla and GoCardless are bank-rail infrastructure with audit-grade APIs, and that's what's commanding strategic premiums, not consumer brand. Expect a SA-flavoured version of this — PayShap orchestration plus card acquiring plus working capital — to be the consolidation play of the next 12–24 months.
Block launched Square AI in Australia as a free conversational assistant embedded in the merchant dashboard — answering natural-language questions about sales, top-selling items, peak hours, year-on-year trends, blended with local context (weather, events). Cited pain point: Australian SMB owners spend 3.34 hours/week on data analysis. Same week, Square added 11-location US chain The Hat to its restaurant book (Q1 Square gross profit +9% to $982m, GPV +13% to $61.2bn).
Why it matters
Square is doing exactly what Toast, Lightspeed, Adyen and Stripe will all do over the next 12 months: turning the dashboard into an AI front-end so that the payment platform owns the merchant's first analytics question. The strategic logic is unchanged from the SaaS playbook — embedded intelligence raises switching costs more than feature additions. For Yoco, iKhokha and the SA bank-acquirer offerings, this sets a clear benchmark: free, conversational, dashboard-native analytics is becoming table stakes, and any acquirer still selling 'reporting' as a paid tier is going to feel margin pressure inside a year. The interesting wedge is that Square's AI works because Square owns the transaction data — which is the same reason Lesaka, Yoco and the major SA acquirers are structurally better placed to deliver this than a horizontal AI vendor selling into the merchant.
Visa launched Agentic Ready in Central Europe, Middle East and Africa with 30+ issuing partners enrolled — including nine UAE banks plus Egyptian, Kuwaiti and Qatari issuers. The programme gives banks a production-grade sandbox to test agent-initiated transactions with tokenisation, behavioural analytics and identity controls layered in. PYMNTS-cited research shows ~60% of UAE businesses are already exploring agentic use cases. Separately this week, Mastercard deepened its CIB Egypt partnership on digital payments.
Why it matters
Two things worth registering. First: Visa is treating CEMEA as a leading market for agentic commerce, not a follower — which is unusual and reflects how mobile-first these markets are. The implication for SA issuers (Standard Bank, Absa, FNB, Nedbank, Capitec, Investec, African Bank, Discovery) is that there's a closing window to be in the first wave of Visa's certification rather than the second. Second: Visa's positioning here aligns with yesterday's GL Insight thesis — networks are competing to be the trust and identity layer for agents, not the settlement rail. If you're a SA fintech building agent-friendly checkout or merchant tech, the integration target is no longer 'add a token vault' — it's 'be on the agent-ready issuer list and support AP2/UCP semantics at acceptance.'
At Biashara Afrika 2026 in Lomé (18–20 May), Afreximbank confirmed intra-African trade hit $220bn in 2024 (+12.5% YoY) and disclosed PAPSS is now live in 21 central banks and 150+ commercial banks with Ghana–Nigeria trades executing in local currency. A $1bn African Collaborative Transit Guarantee Scheme is funded for single-bond movement Cape Town to Cairo. Automotive rules of origin were finalised in February. AfCFTA's ADAPT pilot — Kenya, Morocco, Nigeria — is operationalising shared digital ID and payment-rail infrastructure. But MeTL and Tetra Pak are publicly flagging that digital rails are outpacing physical infrastructure, with 4-month road routes and port capacity still adding billions in cost.
Why it matters
PAPSS at 150+ banks is the moment continental settlement stops being a slide and becomes a routing decision. The honest part of the story is the tension MeTL is flagging: payments infrastructure now moves faster than rail, road and ports, which means cash conversion cycles will improve but inventory financing demand will rise — a clear opening for working-capital fintechs. For SA banks and acquirers, PAPSS interoperability with the SADC-RTGS rail and the upcoming SARB Payments Vision 2030 work is where the next regulatory inflection happens. The strategic question for SA payments operators: do you go direct on PAPSS, or partner with an Afreximbank-anchored aggregator? The pure-play orchestration opportunity here is bigger than the typical 'African expansion' framing suggests.
Pick n Pay sold a further ~12.5% of Boxer for R4.7bn via accelerated bookbuild — its second material Boxer disposal — to fund the core supermarket turnaround now pushed out to 2028 breakeven. The core business carried a R1.5bn trading loss in 2024 and is managing labour disputes touching 22,000 workers. SARS retail data published the same week shows trade sales growth slowing to 2.6% YoY with food and beverage retailers in severe contraction — the Middle East fuel shock we flagged yesterday as compressing 15 million daily commuters' grocery budgets now has a confirming datapoint in the formal retail numbers.
Why it matters
The Boxer disposal pattern is now confirmed structural, not a one-off: Pick n Pay is systematically liquidating its discount-format stake to fund a mid-market rescue. For payments operators, the acquiring mix implications we noted yesterday sharpen: merchant mix is shifting toward discount and informal retail, compressing average ticket while raising transaction count. Pick n Pay's reduced Boxer ownership also weakens the cross-collateralised loyalty case for Smart Shopper at exactly the moment the township/spaza ecosystem — backed by R3bn in Small Business budget — is competing for the same basket.
The agentic commerce stack is now a real procurement decision Google's Universal Cart goes live across Search/Gemini/YouTube/Gmail, Visa's Agentic Ready program enrols 30+ CEMEA issuers (including UAE banks), Fireblocks ships an Agentic Payments Suite via x402, AllUnity launches agentic rails alongside its SEK stablecoin, and Primer raises $100m to make AI the operating layer for orchestration. Merchants now need to decide which protocols (UCP, AP2, ACP, x402) they support — invisibility to agents is a near-term revenue risk.
BNPL keeps migrating from checkout overlay to core infrastructure Worldline×Klarna at acquirer level, Peach×PayJustNow inside SA POS terminals, Klarna's ChatGPT shopping app, Affirm onto Royal Caribbean, and US credit unions building BNPL natively for Gen Z retention. The product is no longer a button — it's a rail. For SA acquirers, the Peach move makes BNPL-at-POS table stakes.
Fintech profitability is being graded harder, even when growth is real Nu Holdings drops 4–6% on a slim revenue miss despite 29% ROE and $5B revenue; Sea and MercadoLibre deliberately compress margins for ecosystem investment; PagSeguro's credit book outpaces deposits as SELIC bites; StoneCo discloses churn from bundling complexity; Paysafe's EBITDA margin slides 130bps on credit losses. Mercury — profitable for four years and now bank-charter-conditional — gets the premium. The market is rewarding capital efficiency, not narrative.
AfCFTA's payment plumbing is shifting from policy to live rails PAPSS is now in 21 central banks and 150+ commercial banks with live Ghana–Nigeria trades; Kenya, Morocco and Nigeria pilot ADAPT; Afreximbank's $1bn African Collaborative Transit Guarantee Scheme is funded. The hard truth from MeTL and Tetra Pak: digital rails outpace physical infrastructure, so the bottleneck is shifting from settlement to logistics.
South African informal and grant economies are being formally digitised on a deadline Cape Town minibus taxis go cashless 1 June; SASSA Gold-to-Black card migration deadline is 31 August; Smart ID expansion to 167 (and onward to 750) bank branches; digital ID regulations gazetted; SARS digital traveller declaration live 1 June. Combined with R302bn Social Development budget, this is a coordinated push to move millions of underbanked South Africans onto digital rails — with real distribution upside for whoever owns the wallet and merchant layer.
What to Expect
2026-06-01—Cape Town minibus taxis go cashless under Codeta mandate — tap cards and SAPAY app only. SARS Traveller Management System goes live for cross-border vehicles.
2026-06-10—Hotel & Hospitality Expo Africa opens in Cape Town (10–12 June) — African hotel AI adoption is reported at 57% vs 35% globally.
2026-06-30—AllUnity targets June 2026 launch of SEKAU, the first MiCAR-compliant Swedish krona stablecoin, alongside its x402-based agentic payments product.
2026-08-31—Postbank deadline for SASSA Gold-to-Black card migration. Service interruption risk for millions of grant recipients past this date.
2026-09-17—Trump executive order's 120-day review deadline for Fed master account access for non-bank fintechs — could materially reshape US settlement economics.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
698
📖
Read in full
Every article opened, read, and evaluated
225
⭐
Published today
Ranked by importance and verified across sources
14
— 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