island fintech jul '26
the grand path of paynow, agentic commerce gets real, SCB tokenizes them deposits
greetings islanders,
I hope you have not missed us too much - the island fintech team have been real busy! we’ve been recording a podcast about what we feel is the future of our industry (stablecoins!) whilst also organising a whole number of fintech runs in the heart of central singapore.
it’s not just this team - southeast asia fintech just had its busiest payments fortnight in years, and it feels like nobody outside the industry noticed?! singapore published the blueprint for PayNow’s next decade and shipped the region’s first governance framework for AI agents that move money. days later, thailand’s siam commercial bank became the first bank anywhere to run live treasury over citi’s tokenised-deposit rails - moving real dollars over a US public holiday, while the fedwire lights were off. not to worry, because we’ve captured it all in this july edition so you can keep afloat of it all.
dips 🎣
1. PayNow gen2: singapore redesigns the rails it already won on
what happened: on 25 june, at the ABS annual dinner, DPM gan kim yong and ABS chairman tan teck long unveiled the phase 1 report of the PayNow generation 2 study - a full rethink of the national instant payments system, built on consultations with 37 organisations and benchmarking against 11 jurisdictions.
four enhancements made the cut:
QR interoperability between PayNow and NETS. today they are separate acquiring schemes - not every banking app can scan a NETS QR, and the NETS app can’t pay a PayNow QR. the fix lets any app pay at any merchant. pilot targeted for end-2026.
deep-linking at online checkout. kills the save-the-QR-screenshot, switch-to-banking-app, upload, return-to-merchant ritual. tap PayNow, your bank app opens with the details pre-filled, approve, done. targeted within a year.
larger-value payments to government agencies, via a sandboxed pilot starting next year. today those run on GIRO, which needs pre-registration and up to three business days to clear.
request-to-pay and structured payment data for automated reconciliation - the business unlock, to be scoped in phase 2. full implementation roadmap due end-2026.
the scale being redesigned: ~11 million proxy registrations (over 90% of adults, ~350,000 businesses), S$154 billion in consumer and S$147 billion in business payment value in 2025.
🏝️ if insight: the deep-linking fix is huge! account-to-account and wallets already hold 48% of retail wallet share in singapore versus a 37% average in card-led markets, and 65% of consulted organisations say PayNow is clearly cheaper than cards. yet cards keep winning online, because habit plus fewer taps beats cheaper acceptance every single time. deep-linking is where PayNow’s cost advantage finally has to pass the UX test.
two more things worth addressing… the phase 1 report explicitly name-checks agentic commerce as a future consideration - the national IPS is being redesigned with software buyers in mind, not just human ones. and if you operate against PayNow scheme rules today, the end-2026 roadmap will reach into your message formats, reconciliation data and fraud handling.
2. singapore writes the rulebook for agents - while the rails go live
what happened: on 3 july, MAS published SAFR (safeguards for agentic finance at runtime), an industry-written white paper under its BuildFin.ai initiative. the premise is blunt: AI agents in finance now execute autonomously at speeds where a human reviewer is decoration. so the controls move to the point of action - runtime checkpoints, think digital tripwires that verify and record what an agent is about to do before it does it, and stop it dead if it strays outside its mandate.
the framework rests on four pillars: policy-bound execution, real-time validation, auditability and interoperability. banks have already applied it to agent-assisted payments and treasury operations, wealth management and client engagement, with the new future of finance institute running pilots and sandbox work.
meanwhile, the commercial rails are being worked on too:
mastercard agent pay is live in singapore and malaysia, with authenticated agentic transactions already run in hong kong and thailand
june brought agent pay for machines - built for continuous agent-to-agent microtransactions, settling across cards, accounts and stablecoins
visa’s intelligent commerce connect lets merchants accept agent-initiated payments across all the competing protocols through one integration, with APAC pilots underway
🏝️ if insight: MAS is betting on industry practice early by setting standards - but standards have a habit of becoming procurement requirements. expect “policy-bound execution” and “auditability at the point of action” to show up in diligence questionnaires within two quarters, whether or not you’ve shipped an agent. cheap to build now, expensive to retrofit later.
on the rails side, ignore the sneaker-buying demos. the B2B story is the money story: sub-cent, always-on, machine-to-machine payments are a description of machine-speed treasury, and a direct challenge to the assumption that cross-border B2B keeps moving in batch files.
we’re converging towards a split stack - card rails for retail, stablecoin and tokenised rails for machine-to-machine and cross-border B2B. the regional winners will be the licensed players who can sit on both sides of that split.
3. tokenised deposits enter live ASEAN treasury
what happened: on 8 july, siam commercial bank became the first financial institution client globally to go live on citi’s newly integrated 24/7 USD clearing and citi token services.
quick decode: citi token services turns ordinary bank deposits into tokens on a private, permissioned blockchain - same regulated dollars, just recorded on rails that never close. 24/7 USD clearing then settles interbank payments as book transfers across the USD accounts that 300-plus financial institutions in 50-plus markets hold with citi, letting money move while fedwire and CHIPS are asleep. until september’s integration, the token service only worked between citi’s own accounts. now it works interbank.
the proof transaction was chosen for maximum symbolism: phillip securities thailand received USD from an affiliate’s citi london account over the US independence day weekend - while conventional US payment systems were closed.
🏝️ if insight: this is tokenisation graduating from pilot theatre to plumbing. weekend liquidity gaps are a real, quantifiable treasury cost, and “money moved on the fourth of july” says more than any white paper.
two caveats before anyone gets carried away. first, this is a single-bank network wearing a “network of networks” costume - you still need an account with citi, so the moat is citi’s balance sheet, not the blockchain. second, it lands squarely on turf stablecoins were supposed to own: always-on cross-border USD settlement. banks are now shipping the same outcome inside the regulated perimeter, with deposit insurance and existing counterparty relationships attached. “stablecoins vs tokenised deposits” is no longer a panel debate - it’s a live procurement decision, and the answer mostly comes down to who your counterparties bank with. either way, client expectations on cut-off times just got reset.
4. indonesia rewrites the payments rulebook (and most PSPs are underestimating it)
what happened: BI regulation 10/2025 took effect on 31 march 2026, and it is the most consequential regulatory change in the region this year that nobody outside jakarta is talking about.
the shift, in brief:
licensing moves from entity-based to activity-based. licences are no longer granted by provider type but by bundles of approved activities. want a new activity? that needs approval.
TIKMI arrives - a mandatory scoring assessment across transaction volume, interconnection, competence, risk management and IT infrastructure. think of it as a credit score for payment companies: it decides whether you’re classified “primary” or “non-primary”, and it now feeds directly into your ongoing capital requirements.
business plans became regulatory documents. every PSP had to file a three-year strategic plan and a one-year payment system business plan by 30 april. partnerships now have to live inside that approved plan.
BI hands down first TIKMI results and classifications by 31 march 2027; non-compliant firms have until 2029 (extendable by two years) to get there.
🏝️ if insight: read the cooperation clause twice, because that’s the one that bites. planned partnerships must be baked into your approved annual plan, and arrangements involving non-indonesian counterparties are classified as “complex”. translation: if you’re a foreign PSP planning to plug into an indonesian partner mid-cycle, the window for doing that quietly has closed.
the second-order effect is consolidation. recalibrated capital thresholds and governance expectations will squeeze the long tail of indonesia’s 500-plus licensed PSPs, and BI almost certainly knows it. indonesia just became a market you enter deliberately, on a 12-to-18-month clock, or not at all. the ASEAN fragmentation problem didn’t get better this year. it got more expensive.
5. the H1 funding numbers look great - until you take the data centres out
what happened: southeast asia’s tech companies raised US$7.4 billion in H1 2026, more than double the US$3.2 billion of H1 2025. sounds like a boom. now the fine print:
funding rounds fell from 153 to 127
the growth is anchored by mega-deals - most notably DayOne’s US$4.5 billion in data-centre raises
fintech specifically dipped 3% to US$685 million
even that “stability” leans on one deal: airwallex’s US$320 million series H at an US$11 billion valuation (up 38% in six months, on annualised revenue of US$1.3 billion)
singapore captured 94% of all regional tech funding, up from 91% in H2 2025
🏝️ if insight: the recovery headline is an infrastructure story wearing a tech costume. strip out the data centres and the picture for fintech founders is capital constraint, later-stage preference, and a bar set at path-to-profitability rather than growth.
two details worth sitting with. first, where the biggest fintech cheque went: airwallex is explicitly raising to push into agentic finance, citing its 85-plus licences and settlement rails as exactly the infrastructure the agent economy needs. capital is following the licence-plus-rails thesis - which should sound familiar if you read dip 2. second, singapore’s 94% share is simultaneously a flex and a failure mode. an ecosystem where nineteen dollars in twenty land in one city isn’t a southeast asian ecosystem - it’s a singaporean one with regional distribution. expect this to sharpen the localisation debate in jakarta, manila and hanoi.
dives 🤿
six things worth your time this fortnight:
paynow generation 2: phase 1 report - MAS/ABS. the primary source. the benchmarking deep dives on malaysia, india, brazil and the UK are quietly the best comparative instant-payments study published this year.
safeguards for agentic finance at runtime (SAFR) - MAS. short, readable, and it tells you exactly what your bank partners will be asking about in six months.
asia’s stablecoin strategy: singapore, japan and hong kong - spark / tiger research. the number that matters: roughly 99% of stablecoins are USD-pegged; local-currency coins are under 1% of the market. everything about asia’s monetary sovereignty anxiety follows from that one line. also the clearest write-up of project BLOOM, the MAS-backed thailand-singapore tokenised settlement corridor.
hong kong’s stablecoin vision - asia stablecoin. the HKMA granted its first two issuer licences in april (HSBC, and the standard chartered-led anchorpoint), with go-live slated between midyear and H2 2026 and 30-plus applications still queued. the beijing subplot around JD and ant is the part most western coverage misses.
indonesia’s updated payment system regulatory architecture - baker mckenzie. the most usable breakdown of PBI 10 and PADG 32. if you have any indonesian exposure, forward this to your compliance lead today.
singapore scam losses and the limits of national action - iPiD. the enforcement machine is genuinely working - scam cases fell 28% in 2025, and a may-june operation with five banks averted another S$38 million. but every win is post-authorisation. there is still no mandated payee name-matching before funds move, and the QR corridors are multiplying faster than any shared verification standard. know-your-payee becomes a procurement requirement before it becomes a regulation.
If you’re building on any of the above - agent controls, indonesian licensing, tokenised settlement - reply and tell me what you’re seeing. the best material in this newsletter usually comes from you.
until next time, island fintech
thanks for reading. feedback, tips or just want to say hi? reach out!