Tokenized equities could change that.

In this episode of WRAP UP, I sit down with Mark Greenberg, Chief Commercial Officer of Kra…

Read more

This week’s Fintech Wrap Up highlights how AI, stablecoins, and modern payment infrastructure are reshaping financial services. AI agents are increasingly taking over payment decisions based on consumer preferences, while stablecoins are gaining traction—particularly for faster, cheaper cross-border payments and AI-enabled financial services. Fintech investment slowed in Q2 2026, with fewer but larger deals, while digital banking funding surged. Meanwhile, banks face growing pressure to replace costly legacy payment systems with agile, API- and cloud-based infrastructure. The report also emphasizes the rise of agentic commerce, where businesses must increasingly optimize not only for human customers but also for AI agents, and highlights JPMorgan’s integrated strategy spanning treasury services, embedded finance, trade finance, merchant services, and digital payment solutions.

Video of the Week

Deep Dive of the Week

When financial institutions should build their own transaction foundation models

Banks and payment companies are changing how they use artificial intelligence. For years, they used separate tools and manual coding to handle tasks like spotting fraud, checking credit scores, and sorting purchases. Building and updating these old systems required a lot of extra work and repeated effort.

Now, companies are switching to larger, unified AI models. Instead of looking at transactions one by one, these new models look at the entire stream of a customer’s activity to understand patterns over time. Financial companies now have to decide if they should build these modern AI tools themselves, buy them from other software providers, or keep tweaking their older, specialized systems.

This week’s reports

1️⃣If AI agents transact autonomously, who controls checkout, payment choice, and loyalty?

2️⃣How Financial Services Leaders Approach The Stablecoin Evolution

3️⃣State of Fintech

4️⃣McKinsey on Finance

5️⃣The Legacy Exit Strategy for Payment Modernization

6️⃣Agentic Commerce: Market to People, Sell to Software

7️⃣Let me break down JPMorganChase’s payment strategy

If AI agents transact autonomously, who controls checkout, payment choice, and loyalty

In traditional e-Commerce, the human shopper makes the decision on which payment method to select based on the options presented by the merchant. Here, the merchant has an element of power to “steer” the consumer towards a certain choice, for example, by offering discounts or pre-selecting a default choice. In the agentic commerce model, the AI agent increasingly acts as the primary decision-maker, choosing the most-preferred payment method that both reflects the consumer’s instructions and is accepted by the merchant.

The consumer will typically specify the following criteria when it comes to the most-preferred payment method:

-Total cost to the consumer

-Available discounts

-Loyalty benefits

It is then up to the agent to pick the payment method that most closely aligns with the specifications above, within the constraints of what the merchant makes available. Over time, with the user’s permission, the agent can also use transaction history data to learn which payment methods are used more in certain contexts and refine the recommendations accordingly.

Read more

Banks and payment companies are changing how they use artificial intelligence. For years, they used separate tools and manual coding to handle tasks like spotting fraud, checking credit scores, and sorting purchases. Building and updating these old systems required a lot of extra work and repeated effort.

Now, companies are switching to larger, unified AI models. Instead of looking at transactions one by one, these new models look at the entire stream of a customer’s activity to understand patterns over time. Financial companies now have to decide if they should build these modern AI tools themselves, buy them from other software providers, or keep tweaking their older, specialized systems.

The architecture of financial grammar

Financial transactions look a lot like human language, but using standard AI models directly on them doesn’t work well in real-world systems. Transactions happen in order, with specific timing, amounts, and results. Turning tabular transaction data into plain text uses up too much memory (about 55 tokens per transaction). With a standard 2,048-token limit, an AI can only look at fewer than 40 past transactions. However, spotting changing spending habits or financial stress requires analyzing hundreds of transactions over many months.

To fix this, companies use compact encoding methods:

• Nubank’s nuFormer: Uses special tokens for numbers and timestamps alongside standard text encoding. This cuts memory use by 75%, allowing the model to look at much longer transaction histories.

• Revolut’s PRAGMA: Converts transactions into structured key-value-time tokens. It groups numbers into ranges, compresses large time gaps while keeping short ones accurate, and tracks daily, weekly, or monthly patterns directly.

Transaction models must also handle two different types of information: static data (rarely changes, like account age) and dynamic data (changes with every transaction, like purchase amounts). Processing both the same way wastes computing power.

Read more

In this episode of WRAP UP, I sit down with Chris Walters, CEO of Finastra, to discuss where AI is creating real value for banks and where the industry needs to slow down.

Chris explains why high-stakes payment decisions still need human oversight, how Finastra’s Global PAYplus platform directs around $7 tril…

Read more

This week’s report explores how AI and digital innovation are transforming the financial sector, from payment modernization and AI-powered financial services to digital currencies and banking automation. It also examines the continued growth of SaaS and fintech companies, the challenges of adapting pricing models to rising AI costs, and the emergence of an economy increasingly driven by autonomous AI agents. Overall, the reports point to a financial industry undergoing rapid technological change, with banks and fintechs needing to modernize their infrastructure, rethink customer experiences, and adapt to new forms of competition.

Video of the Week

Deep Dive of the Week

Structural Logic Behind Stripe’s Acquisition of OpenRouter AI Gateway

Stripe signed a definitive agreement to acquire OpenRouter for a total consideration exceeding $7.5 billion, structured as a combination of cash and stock with a $1.5 billion retention pool allocated for founders and key engineering personnel. The transaction occurred 82 days after OpenRouter finalized a $113 million Series B funding round at a $1.3 billion post-money valuation. The resulting fivefold valuation step-up represents a significant re-rating of artificial intelligence gateway infrastructure within private technology markets.

This week’s reports

1️⃣Partnering for payment modernization

2️⃣Addressing Top of Mind Issues for Banking and Capital Markets

3️⃣The state of intelligent finance in the UK

4️⃣Top banking trends for 2026

5️⃣Q1 2026 SaaS+Fintech Earnings Observations

6️⃣Usage-based Billing: Essential Infrastructure, Wrong Buyer Interface

7️⃣Machine Economy Rising: The Market of Limitless Cognition

Partnering for payment modernization

Key findings (Based on a survey of 500 bank and 500 retail payment executives conducted by KPMG International)

Partnerships

With 53 percent of retailers saying that their banks understand their payment modernization goals and 51 percent of banks saying that the future winners in payments will be those with the best ecosystems, partnerships emerge as a key success factor in payment modernization.

Customer expectations

Banks and retailers say one of their top motivations for payment modernization is to meet customer demand for seamless, convenient, and trusted payment options, but only half of retailers say their banks are helping them improve the customer payment experience.

Innovation

Read more

Stripe signed a definitive agreement to acquire OpenRouter for a total consideration exceeding $7.5 billion, structured as a combination of cash and stock with a $1.5 billion retention pool allocated for founders and key engineering personnel. The transaction occurred 82 days after OpenRouter finalized a $113 million Series B funding round at a $1.3 billion post-money valuation. The resulting fivefold valuation step-up represents a significant re-rating of artificial intelligence gateway infrastructure within private technology markets.

Prior to the acquisition, OpenRouter was processing over 200 trillion tokens monthly, translating to roughly 10 trillion daily tokens across 10 million registered software developers and enterprise accounts. The network aggregates access to 500 individual foundation models across 80 underlying inference providers.

The speed of this acquisition cycle reflects a broader shift in institutional capital deployment. Strategic buyers and venture backers re-evaluated the platform as a core utility layer for artificial intelligence traffic clearing.

Token clearinghouse economics

The strategic foundation of this transaction rests on the premise that tokens have become a primary unit of account and medium of economic value exchange for modern software applications. While foundation model developers engage in intense price competition that compresses compute margins, total inference volume continues to expand exponentially. Business spending telemetry across enterprise software accounts indicates token consumption grew thirteenfold year-over-year.

OpenRouter extracts a 5.5% fee on token purchases routed through its marketplace, alongside tier-based monetization for bring-your-own-key configurations. Operating at the network routing layer insulates the platform from model-level

Read more

In this episode of the WRAP UP, I sit down with Michael Blau, founder of Drip, to di…

Read more

What does stablecoin compliance look like in 2026? We are all busy with technology and how great it is, but without the right compliance, stablecoins won't work.

Between 2023 and 2026, the GENIUS Act, MiCA, the UK's FCA regime, and dedicated frameworks from MAS, the HKMA and VARA all came into force on top of a rewritten FATF Travel Rule.

Now the question is which rules apply to your specific activity, and that rarely has a simple answer. A bank or a payments company can all touch the same stablecoin and carry completely different obligations, depending on their role and stacks.

If you want to understand what compliance frameworks apply to your business, Sumsub's report helps to clarify:

- What the seven frameworks that matter — GENIUS, MiCA, the FCA regime, MAS, HKMA, VARA, and the FATF Travel Rule — actually require, and where each one stands today
- How obligations differ depending on what you do: issuing, custody, payments, wallets, banking, or card programs
- The common pitfalls compliance teams hit when stablecoin activity gets bolted onto an existing program instead of built into it
- A practical readiness checklist for each business segment, so you can test whether your controls are keeping pace with regulatory expectations

You'll also see where the rules are trending. FATF Travel Rule adoption has reached 83% of surveyed jurisdictions, up from 73% a year earlier. Stablecoins' share of crypto transfers has climbed from 31% to 36% over the same period. The scrutiny has shifted too, from checking whether a policy exists on paper to testing whether controls actually work in practice. This guide shows what that bar looks like for your segment.

Read The Report

Video of the Week

Deep Dive of the Week

The AI Model Gateway/Router Directory For Fintechs and Banks

Stripe just acquired OpenRouter for more than $7 billion, roughly five times the valuation OpenRouter carried three months ago. A payments company just paid a premium to own the layer that decides which LLM handles a request, at what cost, and with what fallback path. If you build fintech products on top of LLMs, this is no longer a developer tooling question. It is an infrastructure decision with the same weight as picking a payments processor or a core banking vendor.

Stripe’s acquisition of OpenRouter turns AI gateways from a nice-to-have into infrastructure banks and fintechs need to evaluate seriously. This guide maps 16 AI gateway and LLM routing platforms across five categories, with the deployment models, compliance posture, and pricing structures that matter when you handle regulated data. I built it because most comparisons of this category are written for general SaaS teams, not for people who need to answer a compliance questionnaire before they ship. Structured profiles and a companion Excel file with seven filterable tabs are below the paywall. Use it to shortlist vendors before your next infrastructure review.

This week’s reports

1️⃣The Next Payment Infrastructure in Africa

2️⃣AI and Stablecoins’ Transformation of the Global B2B Payments Experience

3️⃣UAE Digital Assets and Stablecoin Ecosystem

4️⃣Stablecoin playbook

5️⃣Q2 2026 Quarterly FinTech Insights

6️⃣Revolut - The Dawn of Modern Banking

7️⃣Stablecoins in Europe

The Next Payment Infrastructure in Africa

Africa has built one of the world’s most dynamic financial ecosystems by leapfrogging traditional banking infrastructure directly into mobile-first networks. Driven by a young median age of 20 and rapid urbanization, domestic demand for digital services, e-commerce, and merchant payments continues to accelerate.

However, domestic digital success has not created a unified continental ecosystem. Payment adoption remains deeply uneven — spanning from over 70% digital payment usage in Senegal to under 10% in Niger. For businesses, domestic efficiency stops at national borders, exposing a stark contrast between fast local mobile transactions and complex, costly cross-border commercial settlement.

Four Frictions Limiting Payment Efficiency

While domestic Instant Payment Systems (IPS) expanded to 33 active networks generating 65.6 billion transactions in 2024, cross-border business payments remain constrained by severe structural inefficiencies:

• The Formalization Gap: With 85.8% of African employment situated in the informal economy, smaller enterprises and merchants struggle to access standardized formal banking and payments.

• Payment Rail Fragmentation: Incompatible technical and regulatory frameworks across national borders restrict direct payment routing and degrade cross-border success rates.

• FX and Liquidity Traps: Currency fragmentation forces intra-African commercial transactions through foreign currencies and overseas correspondent banks, generating an estimated $5 billion annually in unnecessary transaction costs.

• Settlement Latency: Multi-hop correspondent banking pathways increase processing friction, resulting in high fees and settlement delays spanning multiple business days.

Stablecoins as a Value Movement Layer for Global Trade

To overcome traditional banking friction, commercial enterprises across Africa are turning to alternative blockchain-based settlement mechanisms. Between July 2024 and June 2025, Sub-Saharan Africa received approximately $205 billion in total on-chain value (+52% YoY).

Dollar-denominated stablecoins (such as USDC and USDT) are increasingly integrated into real-world trade flows between Africa, the Middle East, and Asia — including multi-million-dollar energy and merchant settlements. By routing value across blockchain networks into local off-ramps, stablecoin pathways reduce correspondent banking bottlenecks, secure dollar liquidity, and compress cross-border settlement from days to minutes while lowering costs from $5–$15+ to under $1.

Report by Photonpay

AI and Stablecoins’ Transformation of the Global B2B Payments Experience

Artificial intelligence is by no means new to payments. Earlier forms of the technology have played a role in cross-border transactions for some time, particularly in areas such as know-your-customer (KYC) and fraud mitigation. But recent developments have progressed the technology considerably.

The evolution of generative AI and large language models (LLMs) such as ChatGPT and DeepSeek initially sparked a surge of hype across payments and beyond. Boardrooms around the world were seized with discussion on how the technology could help their bottom line – as well as how it might erode their well-honed business models.

The technology was subject to a lot of excited discussion, but also a lot of noise and exaggeration. Now the hype has reduced, the actual benefits are easier to see – and for business payments, the use cases are significant.

AI’s Payments Impact: Numbers from the Industry

+9-15% Mastercard has seen from initial trials of its AI-based Payment Optimization Platform for merchants

-75% support times money transfers provider Remitly saw when it launched its AI-powered virtual assistant

-30% The drop in fraud losses from card scams that Revolut saw after it introduced a new AI-based feature to detect if a customer was being scammed

+200% efficiency SUNRATE saw from its internal AI tool

$3tn-$5tn The opportunity that J.P. Morgan sees in “services as a software” – the delivery of business services via AI-powered software platforms

50+ The number of use cases Swift is testing or using AI for, including fraud detection, information sharing and efficiency improvements

Report by FXC Intelligence and Sunrate

UAE Digital Assets and Stablecoin Ecosystem

Five years after establishing VARA as the world’s first dedicated virtual asset regulator, the UAE’s digital asset ecosystem has moved from framework-building into regulated commercial execution.

The May 2026 MENA Fintech Association roundtable at Daos Hub reflected this shift: discussion centered not on whether to regulate, but on converting existing infrastructure into economic activity at scale. AED stablecoin initiatives have moved toward operational use. VARA-licensed VASPs now number 50, with the regulator targeting 80% by year end. On May 12, the UAE approved crypto for government fee payments. DDSC’s dirham-backed stablecoin ecosystem (IHC, First Abu Dhabi Bank, Sirius) received Central Bank approval, and IHC executed a ~AED 110 million (~USD 30 million) transfer on ADI Chain — one of the UAE’s largest disclosed stablecoin transactions. The UAE’s positive FATF standing further reinforced compliance confidence.

Three themes emerged. First, stablecoins are now operational reality: global on-chain settlement volume hit a gross USD 33 trillion in 2025 (75% YoY growth), surpassing combined Mastercard/Visa volumes — though bot-adjusted figures put genuine economic activity closer to USD 9 trillion. The AED stablecoin is live domestically, but cross-border adoption lags due to onshore custody requirements and unstandardized FX swap workflows.

Second, banking access remains a core operational challenge — licensing alone doesn’t resolve the need for operating accounts, client money accounts, and correspondent banking, particularly for firms serving multiple jurisdictions.

Third, global competition has intensified. The UAE’s early-mover advantage is being tested as the US (GENIUS Act enacted, Clarity Act advancing) and EU (MiCA established) build clearer frameworks. Going forward, UAE’s edge will depend less on being first and more on ecosystem depth and commercial effectiveness.

The foundation is strong; the next phase hinges on scaled commercial use, deeper bank participation, cross-border settlement, and coordinated infrastructure — converting regulatory leadership into durable market leadership.

Insights by MENA FINTECH ASSOCIATION

Stablecoin playbook

The global financial system is standing at the threshold of a fundamental reset. Stablecoins, digital assets whose value is pegged to a stable reference, most often a single fiat currency such as the US dollar, have emerged as a critical foundation for the new financial infrastructure. These fiat-linked stablecoins, fully backed by cash or cash equivalents and designed for minimal price volatility, offer the trust and stability required for mainstream adoption.

Financial products are transitioning on-chain, with tokenized versions of regulated assets (e.g., stocks, funds, deposits) now operating on public blockchain infrastructure. The transition is shifting how value is stored, transferred, and programmed. Stablecoins have evolved from simple trading pairs for crypto markets to becoming the backbone of programmable money. This shift toward a digitally native financial system is explored further in “From vaults to virtual: Digital revolution in the financial system.” Forward-thinking institutions now see stablecoins as strategic solutions to long-standing challenges in cross-border payments, treasury operations, and liquidity management. Increasingly, stablecoins serve as the foundational settlement layer for a tokenized financial ecosystem.

As of April 2026, the stablecoin market is valued at approximately US$316 billion, with Citi projecting issuance of US$1.9 trillion (base case) to US$4 trillion (bull case) by 2030. This exponential growth signals digital-dollar infrastructure maturing into mainstream financial plumbing, driven by institutional demand, the need to address correspondent banking friction, and the rise of programmable payment rails. This transformation aligns with long-awaited regulatory clarity, as the US Guiding and Establishing National Innovation for US Stablecoins Act (GENIUS Act), the EU Markets in Crypto-Assets Regulation (MiCA) framework, and emerging licensing regimes establish robust compliance pathways around the globe.

US dollar-denominated stablecoins currently dominate existing issuance and usage. Whether this dollar dominance persists will depend on regulatory clarity and market structure outside the US, demand for local-currency settlement, and the emergence of credible alternatives (e.g., euro or other major-currency stablecoins and tokenized deposits) that can match USD stablecoins on trust, liquidity, and interoperability, particularly in regional payment corridors and multicurrency corporate treasury use cases.

Report by Deloitte & Ava Labs

Q2 2026 Quarterly FinTech Insights

Key Insights:

- Public Markets & IPO Pause

IPO Slowdown: Public-market activity dropped sharply following Q1’s five U.S. IPOs ($2.0B raised).

Overall Volume: Total FinTech deal volume fell for the 3rd consecutive quarter to $42.2B (lowest since Q4 2024).

International Listings: International IPO activity was minimal, totaling ~$0.2B combined from OnEMI Technology Solutions and Sunmi.

- Private Capital & Mega-Rounds

Private Funding Rebound: Private financing rose to $17.7B (vs. $16.1B in Q1 2026 and flat YoY vs. Q2 2025).

Capital Concentration: Growth was heavily skewed toward six $500M+ megadeals ($4.2B volume). Excluding these, underlying financing declined.

Deal Count Drop: Overall deal count fell 14% Q-o-Q to 696 deals (down 29% YoY). However, 43 deals topped $100M+ (highest quarterly total since Q3 2022).

Stage Breakdown:

Seed & Series A: Stable at 230 deals ($2.8B).

Series B+: Up to 74 deals ($7.8B) vs. $6.7B in Q1.

Top Sector & Investors: Financial Management Solutions led early and late stages. Andreessen Horowitz led VC activity (22 new, 34 total H1 deals). Coinbase Ventures and Tether were top strategic investors. Corporate participation hit 37% overall and 56% in Crypto/Blockchain.

- Unicorns & Major Financing Rounds

New Unicorns (Q2): Corgi ($2.6B), D360 ($1.6B), KreditBee ($1.5B), Slash ($1.4B), nesto ($1.05B), and Farther ($1.0B).

Valuation Increases: Ramp ($44B), Airwallex ($11B), AlphaSense ($7.5B), Alan ($6.3B), and Mercury ($5.2B).

Largest Rounds:

CRED ($900M Series H)

Ramp ($750M Series F)

Ebury ($742M Strategic)

Alan ($550M Series G)

Clip ($500M Late-Stage)

- M&A Activity ($24.3B Total Volume)

M&A stepped down from 2025’s pace with zero $15B+ transactions, but strategic acquirers drove key infrastructure consolidation:

Bullish \ Equiniti ($4.2B): Market infrastructure meets tokenization.

Nuvei \ Payoneer ($2.75B): Cross-border payments and stablecoin-enabled commerce consolidation.

Wafra \ Navitas ($1.9B): Sponsor appetite for specialty finance.

Other Notable Deals: Adyen / Talon.One ($880M), PayPay / T&D Financial Life ($840M), CoStar / Zonda ($800M), Amex / TheFork ($700M), Kraken / Reap ($600M), and Kraken / Bitnomial ($550M).

Report by FTTechnologies

Revolut - The Dawn of Modern Banking

Key insights:

For the love of product.

Product is the growth engine. Every product must “deliver WOW”, and customers reciprocate: around two-thirds of new customers arrive organically or by referral, and every new vertical launches into an already engaged base at minimal acquisition cost, creating a flywheel where each product deepens retention and funds the next. The same discipline now extends to AI, with AIR putting the entire ecosystem behind a single conversational interface.

Complexity that feels simple.

Rivals, at best, focus on product breadth or monetisation depth; Revolut is successfully doing both, distributing institutional-grade, high-margin products to a 75-million-user base through the same daily habit that began with free FX. Three moats - distribution, in-house technology and the proprietary data those users generate - reinforce one another, and the combined advantage compounds with every launch.

Sustainable product velocity.

Each product launch reaches milestones faster than the previous one. New verticals inherit a base that is already at scale and infrastructure that the earlier products had paid for. Payments took eight years to mature, wealth took six, RevPoints is moving faster still. In the UK, the world’s most competitive fintech market, a new core vertical roughly every 18 months took Revolut from zero to 13 million customers by 2025.

Standardise to scale, localise to win.

One playbook, 48 countries. Revolut runs a single global technology stack and brand, but calibrates market entry, product sequencing and licensing to each geography. Having hard-won its home base in Europe, Revolut is now replaying that playbook across Latin America. With the biggest prize, the US, next in line.

Modular by design.

The legal architecture mirrors the tech stack: each business line independently capitalised, separately supervised, free to grow without constraining the others. The structure has earned top-tier standing: ECB supervision in the EU, full UK authorisation completed March 2026, and a US national bank charter filed. The domino effect has started.

Culture rooted in excellence.

Ownership, grit and a relentless pursuit of being number one. Hiring is treated as an engineering problem: 1.7 million applications in 2025, and only 0.15% accepted (comparable to the likes of Google and Goldman Sachs). The talent density shows up outside the company too, with Revolut now one of Europe’s most prolific founder factories, its alumni companies having raised over $1.2bn.

Financial outlook.

Best-in-class financials at scale. $6.0bn of revenue in 2025, up 46%, with $1.7bn net profit at a 29% margin - the fifth consecutive profitable year, on headcount up just 10%. And at only $86 of revenue per customer, monetisation is still largely uncaptured: we project a $29bn topline and $13.6bn of net profit by 2030.

Report by Bullhound Capital

Stablecoins in Europe

Three forces have converged in the last year and a half: the geopolitical environment has made stablecoins strategically necessary, the legal framework has gone from theoretical to operational, and the unit economics now beat the incumbent rails.

The geopolitical environment has changed. With the war in Ukraine entering its fifth year, the dollar openly weaponized through sanctions, and tariff escalation between the US, China and the EU, holding USD on a US bank rail is no longer a neutral choice for European corporates: it carries jurisdictional risk. In emerging markets where local currencies have lost 20 to 60% versus the dollar (Turkey, Argentina, Egypt, Nigeria), stablecoins have become a plausible savings instrument. The result is a structural pull on euro-denominated, MiCA-compliant stablecoins as a sovereign alternative to USD on-chain.

MiCA is no longer a theoretical framework, but a fully established system. Two years in, the first cohort (Circle EURC, SG–Forge EURCV, Monerium EURe, Angle agEUR) has been joined by AllUnity EURAU, StablR and Quantoz, all operating under EMI licenses from BaFin, the ACPR, the AFM and the CSSF. EUR stablecoin supply has grown roughly 5x since 2024, and for European corporates a euro stablecoin is now treated the same way as a euro IBAN. The US caught up in 2025 with the GENIUS and STABLE acts.

The use-case economics have flipped. Settlement on Ethereum, Solana, Base and Stellar is fast and cheap enough that a €50 cross-border transfer costs cents and settles in seconds. Tether, Circle and Paxos collectively manage over $250B in reserves yielding ~4.5%, which fuels their aggressive product expansion. On real corridors (EU–MENA, EU–LATAM), B2B settlement now clears at 30 to 80 bps versus 200-400 bps through correspondent banking. For a treasurer running weekly payroll, the switch pays for itself in one cycle.

Report by Xange

Thanks for reading Fintech Wrap Up!

Disclaimer:

Fintech Wrap Up aggregates publicly available information for informational purposes only. Portions of the content may be reproduced verbatim from the original source, and full credit is provided with a “Source: [Name]” attribution. All copyrights and trademarks remain the property of their respective owners. Fintech Wrap Up does not guarantee the accuracy, completeness, or reliability of the aggregated content; these are the responsibility of the original source providers. Links to the original sources may not always be included. We use AI tools to brainstorm, draft, edit, and summarize, but the ideas, analysis, and connections Sam draws are his. Every piece is verified and written in Sam’s voice. For questions or concerns, please contact us at sam.boboev@fintechwrapup.com.

Stripe just acquired OpenRouter for more than $7 billion, roughly five times the valuation OpenRouter carried three months ago. A payments company just paid a premium to own the layer that decides which LLM handles a request, at what cost, and with what fallback path. If you build fintech products on top of LLMs, this is no longer a developer tooling question. It is an infrastructure decision with the same weight as picking a payments processor or a core banking vendor.

Stripe’s acquisition of OpenRouter turns AI gateways from a nice-to-have into infrastructure banks and fintechs need to evaluate seriously. This guide maps 16 AI gateway and LLM routing platforms across five categories, with the deployment models, compliance posture, and pricing structures that matter when you handle regulated data. I built it because most comparisons of this category are written for general SaaS teams, not for people who need to answer a compliance questionnaire before they ship. Structured profiles and a companion Excel file with seven filterable tabs are below the paywall. Use it to shortlist vendors before your next infrastructure review.

What this guide covers

I collected 16 companies in the AI gateway and LLM routing space, pulling data directly from official websites and public documentation. For each one I looked at what they do, how they deploy (SaaS, self-hosted, VPC, on-prem, hybrid), which LLM providers they support, their feature set across 13 production-relevant dimensions, their compliance posture (SOC 2, ISO 27001, GDPR, HIPAA, data residency), and their pricing model. Where a company did not publicly disclose something, I marked it “Not disclosed” rather than guess. The Excel file breaks this into seven tabs: Overview, Deployment, LLM Providers Supported, Features, Compliance and Security, Pricing and Funding, and a Master Table for filtering across every dimension at once.

The 16 companies, by category

General-purpose LLM routers

OpenRouter routes across 80-plus providers and 500-plus models, now under Stripe. Inworld Router uses conditional logic to route based on user context, zero markup. Unify AI has shifted toward agent workflows but still routes across major providers.

Enterprise and compliance-focused routers

Kong AI Gateway extends Kong’s API gateway into LLM and agent traffic. TrueFoundry AI Gateway supports VPC, on-prem, and air-gapped deployment. Nexos ai runs out of Europe with ISO 42001 and zero data retention options.

Cost-optimization routers

Not Diamond predicts the optimal model per request to cut inference cost while holding or improving accuracy, positioned heavily toward coding agents.

Open-source and self-hosted routing frameworks

LiteLLM is the most widely adopted open-source gateway. Maxim Bifrost offers a high-throughput OSS gateway plus enterprise option. LLM Gateway is AGPLv3-licensed with SaaS, self-hosted, and hybrid options. KrakenD extends its API gateway into AI routing with a free community edition.

AI gateway platforms with routing

Portkey combines a SaaS control plane with an optional self-hosted gateway across 1,600-plus models. Braintrust is primarily observability and evaluation with gateway-adjacent capability. Helicone pairs routing with LLM observability. Cloudflare AI Gateway runs on Cloudflare’s network with dynamic routing. Vercel AI Gateway offers zero-markup pricing and routes coding agent traffic too.

What’s behind the subscribe button

Subscribers get the full structured profile for all 16 companies, organized by category, covering what each company does, its exact deployment model, which LLM providers it supports, a feature-by-feature breakdown across 13 production-relevant capabilities, its compliance and security certifications, and its pricing model. You also get the companion Excel workbook with seven tabs for side-by-side comparison and filtering, plus a closing section on how to actually use this guide when running a vendor evaluation.

If you are choosing infrastructure that will sit between your product and every LLM call you make, subscribe to get the full comparison and the spreadsheet.

Read more

In this episode of the WRAP UP, I sit down with Martin Della Chiesa, CEO of Skaleet, to discuss why replacing legacy core banking systems remains so difficult and how cloud-native, modular and real-time infrastructure is changing the m…

Read more

This week’s reports focus on how AI, data, and new financial infrastructure are reshaping banking and commerce. Key themes include transaction foundation models as a shared intelligence layer for fraud, credit, and payments; non-banks increasingly capturing parts of SMB banking relationships; the growing need for robust on-chain data infrastructure; and the role of Open Finance as a foundation for AI-powered financial services. Other reports highlight payments intelligence and agentic commerce as emerging competitive advantages, while Evident’s LATAM AI Index shows Brazilian banks—led by Nubank and Itaú—outpacing the region in AI adoption and measurable use cases, despite talent and ROI challenges.

Video of the Week

Deep Dive of the Week

The US Expense/Spend Management Companies Directory

I built a structural map of the US spend management category because the space has quietly split into distinct business models that get lumped together under one label. Some of these companies are regulated issuers running their own card programs. Others route every card through a partner bank and never touch a BIN. This guide breaks down 22 platforms across five segments, with license type, card networks, card types, features, and regional coverage for each. I also built a companion Excel file with the full dataset so you can filter and compare without scrolling through prose.

This week’s reports

1️⃣How to build transaction foundation models in banking and payments

2️⃣Non-banks are winning business banking

3️⃣The Gateway to Digital Asset: On-Chain Data Infrastructure

4️⃣Artificial intelligence and open finance

5️⃣The next era of commerce will be won on payments intelligence

6️⃣Agentic Commerce: China’s Lead, Europe’s Choice

7️⃣AI in Index in Banks - LATAM

How to build transaction foundation models in banking and payments

Transaction foundation models (TFMs) are emerging as a new intelligence layer for banking and payments. They are designed to learn from transaction histories such as payment events, customer behavior, merchant activity and operational outcomes, and turn that learning into reusable embeddings that can improve multiple downstream use cases.

The idea is simple but powerful.

Instead of building a separate feature pipeline and model for every problem (fraud, credit, authorization, reconciliation, customer engagement, liquidity forecasting), an institution can build one shared representation of transaction behavior and reuse it across many models and workflows.

Read more

I built a structural map of the US spend management category because the space has quietly split into distinct business models that get lumped together under one label. Some of these companies are regulated issuers running their own card programs. Others route every card through a partner bank and never touch a BIN. This guide breaks down 22 platforms across five segments, with license type, card networks, card types, features, and regional coverage for each. I also built a companion Excel file with the full dataset so you can filter and compare without scrolling through prose.

What this guide covers

I looked at 22 companies spanning US SMB and mid-market card platforms, US spend management tools, enterprise T&E and spend suites, AP automation platforms with card layers, and travel and expense tools. For each one I pulled license type and issuing jurisdiction, card networks supported, card types offered, core feature set, and regional coverage, sourced from official sites and public disclosures. Where a company does not disclose something, I marked it “Not disclosed” rather than guess.

The accompanying Excel file has seven tabs: Overview, Licenses, Card Networks, Card Types, Features, Regions, and a Master Table that puts every company against every dimension in one filterable view.

The 22 companies

US SMB and mid-market card and spend platforms

Ramp issues Visa corporate cards through a multi-bank program and automates expense coding with AI, aimed at fast-scaling SMBs and mid-market companies. Brex runs corporate cards and a broader finance stack through undisclosed program banks, built for startups through enterprise. Mercury layers cards and bill pay on top of FDIC-insured business banking through Choice Financial Group and Column N.A. Rho combines banking, cards, and bill pay through Webster Bank as an alternative to stitching separate vendors together. Slash issues Visa cards directly through Column N.A. with tiered programs tied to underlying balances. Light is an agentic accounting and ledger platform for multinational finance teams, not a card issuer.

US spend management platforms

Airwallex runs global multi-currency accounts, corporate cards, and payments under one API, issuing through Community Federal Savings Bank and Evolve Bank & Trust. Expensify pairs its own Visa Commercial Card with a bring-your-own-card model reconciling against 10,000-plus banks. Bill.com built its footprint on AP and AR automation before adding the Divvy-branded card as a spend entry point.

Enterprise T&E and spend platforms

Navan unifies travel booking with corporate cards and expense automation, issuing through Celtic Bank in the US and Stripe and Adyen entities abroad. SAP Concur is the dominant enterprise travel, expense, and invoice suite, a software layer over whatever card program a customer already runs. Pleo issues Mastercard-branded company cards from its own FCA-authorized e-money entity, mostly serving European finance teams. Payhawk delivers cards, expenses, AP, and procurement as a Visa principal member with UK and EEA EMI licenses plus Cross River Bank in the US. Spendesk unifies cards, expense reports, AP, and procurement through its own regulated payment institution in France.

AP automation and spend platforms

Tipalti automates global mass payouts, AP, procurement, and expenses across 200-plus countries without issuing its own cards. Coupa runs enterprise procurement and total spend management for more than half of the Fortune 500, also without card infrastructure. Stampli layers a card product onto an ERP-aligned procure-to-pay platform. N2F runs expense reports, corporate cards, and AP as three modules across 86 countries, mostly in Europe.

Travel and expense platforms

Rydoo covers 130-plus countries with localized tax and per diem data, issuing its own cards without disclosing the underlying network. Fyle, now Sage Expense Management, reconciles directly against existing Visa, Mastercard, and Amex cards rather than issuing its own. Zoho Expense is a pure SaaS expense and travel tool with no card issuance, pulling feeds from Visa, Mastercard, Amex, HSBC, and AirPlus. Webexpenses pairs a mature expense-report core with an add-on Mastercard cashback card, mostly serving UK and APAC customers.

What’s behind the subscribe button

Subscribers get the full structured profile for each of the 22 companies: what they do, license type and issuing jurisdiction, card networks, card types, key features, and regional coverage. That includes the underlying issuing entities most of these platforms don’t put on their homepage (Celtic Bank, Column N.A., Stripe, Adyen, Cross River, Sutton Bank, Evolve, and others), plus the companion Excel file with all 22 companies mapped across every dimension in a filterable master table.

Subscribe to get the full breakdown and the Excel file.

Read more

Jess Houlgrave, CEO of WalletConnect, joins me to discuss where stablecoin payments make sense and what is holding them back.

WalletConnect works with around 700 wallets and 80,000 applications, reaching about 800 million end users. Around $400 billion in value moved throu…

Read more

The reports highlight major shifts across fintech, with AI governance, tokenisation, stablecoins, CBDCs and agentic payments moving from experimentation toward wider adoption. Banks are increasing AI use but still face governance and risk gaps, while tokenisation could streamline asset markets through automation and more efficient settlement. Stablecoins are gaining regulatory clarity and appear most valuable in markets with inefficient cross-border payment infrastructure, where they are likely to complement rather than replace fiat rails. Meanwhile, China’s e-CNY 2.0 demonstrates how CBDCs can integrate with commercial banks, and the rise of agentic payments points toward AI systems increasingly executing B2B transactions autonomously under controlled frameworks.

Video of the Week

Deep Dive of the Week

Engineering the Agentic Control Orchestration in Banking

Banks often fail to derive meaningful value from AI because they tend to layer sophisticated automation on top of rigid, inefficient legacy workflows. While billions are spent on advanced tools, the actual impact remains minimal as these applications merely mask manual bottlenecks like compliance and security checks rather than resolving the underlying structural flaws. Consequently, achieving true operational progress depends on fundamentally redesigning core processes to integrate with AI, rather than simply deploying assistants to accelerate outdated habits.

To convert localized productivity into a scalable organizational resource, banking institutions must transition to governed agentic work. This shift requires a centralized control plane that decouples core business logic from individual model providers. This article details the technical implementation, system topology, and security parameters of an end-to-end agentic operating platform, utilizing Zafin AIOS as the architectural reference model.

This week’s reports

1️⃣Banking on trust: AI governance for growth, resilience and scale

2️⃣Tokenised Assets

3️⃣AI and Stablecoins’ Transformation of the Global B2B Payments Experience

4️⃣China’s Central Bank Digital Currency 2.0

5️⃣The global expansion of stablecoins

6️⃣Defining Agentic Global Payments

Banking on trust: AI governance for growth, resilience and scale

Banks are accelerating AI initiatives, but governance is struggling to keep pace. Deloitte Access Economics’ latest research shows that stronger AI governance is linked to broader AI deployment and higher revenue growth, positioning governance as a growth enabler rather than a brake on innovation.

As adoption moves from experimentation to enterprise-wide deployment, banks are unlocking efficiency gains and new growth opportunities, while also facing increasing data, regulatory, cyber and reputational risks.

In this paper, Deloitte benchmarks AI governance maturity across Global and Domestic Systemically Important Banks (G-SIBs and D-SIBs), as well as other large banks, using Deloitte’s Trustworthy AI Governance Index. The index provides a snapshot of how global banks are progressing across key dimensions, including principles and policies, organisational structure, procedures and controls, people and skills, and monitoring, reporting and evaluation.

The findings suggest that while progress has been made, most banks still have significant room to strengthen their governance frameworks. The research also explores the link between AI governance and revenue outcomes and outlines practical steps for leading governance transformation.

Key takeaways 

Read more

Banks often fail to derive meaningful value from AI because they tend to layer sophisticated automation on top of rigid, inefficient legacy workflows. While billions are spent on advanced tools, the actual impact remains minimal as these applications merely mask manual bottlenecks like compliance and security checks rather than resolving the underlying structural flaws. Consequently, achieving true operational progress depends on fundamentally redesigning core processes to integrate with AI, rather than simply deploying assistants to accelerate outdated habits.

To convert localized productivity into a scalable organizational resource, banking institutions must transition to governed agentic work. This shift requires a centralized control plane that decouples core business logic from individual model providers. This article details the technical implementation, system topology, and security parameters of an end-to-end agentic operating platform, utilizing Zafin AIOS as the architectural reference model

From individual assistants to governed systems

The integration of artificial intelligence inside financial institutions moves through three distinct operational phases, beginning with humans operating alongside generic chat interfaces for localized productivity lifts. This evolves into automated task pipelines where agents execute processes with human-in-the-loop validation to manage escalating coordination overhead, ultimately reaching a mature state of fully governed operations. In this final phase, autonomous agents manage complete end-to-end system lifecycles while human focus shifts to setting policies, reviewing systemic exceptions, and validating immutable audit trails.

In Phase 1, humans operate alongside generic chat interfaces or local code completion utilities. This pattern yields a localized, non-scalable productivity lift.

Phase 2 introduces automated task pipelines where agents execute processes with human-in-the-loop validation. This architecture delivers exponential productivity gains but introduces coordination overhead as the number of parallel tasks escalates.

Phase 3 establishes fully governed operations. In this mature state, autonomous agents manage complete end-to-end system lifecycles while humans focus on setting policies, defining boundaries, reviewing systemic exceptions, and validating immutable audit trails.

Deploying Phase 3 operations in highly regulated banking environments requires an orchestration platform that acts as an enterprise operating system. Zafin AIOS implements this conceptual kernel, solving the structural vulnerabilities of ad-hoc multi-agent fleets: context fragmentation, scheduling conflicts, lack of environment isolation, and inconsistent tool input/output protocols.

Architecture of the agentic orchestrator

The Zafin AIOS architecture functions as an operating system kernel that sits between the developer-facing application layer and the underlying model endpoints, database engines, and software-as-a-service platforms.

The system organizes agentic work through five primary control layers.

1. The core layers

The platform coordinates operations by dividing system responsibilities into dedicated structural domains.

• Work layer (Agent Work Management): Manages the complete lifecycle of tasks, moving cases through multi-step workflows, tracking active states, and persisting execution context.

• Agents layer (Agent Config Management): Configures and manages the profiles, permissions, compute constraints, model guidance configurations, and identity parameters of synthetic workers.

• Compute layer: Provisions and manages isolated physical or virtual compute runtimes across public cloud, private cloud, or on-premises server infrastructure for agent execution.

• Control layer: Manages continuous system orchestration, real-time logging, LLM traces and audit logs including prompts, inputs, and outputs, token and compute cost tracking, and work item analytics.

• Evaluations layer: Executes automated testing, benchmark comparisons, drift detection, knowledge poisoning detection, and LLM-as-a-judge review cycles.

2. The internal agent kernel structure

Within Zafin AIOS, a running agent instance functions as a structured runtime containing dedicated service interfaces rather than a simple script calling a public model API.

Connected directly to the Platform Agent Config Management database to sync current system rules, the Agent Instance houses a core AIOS Agent Orchestrator. The AIOS Agent Orchestrator requests work from the AIOS Agent Work Management module when it is idle.

Communication with external platforms occurs through a Cross-Harness Interface, which allows engineers to plug in optimized agent harnesses like Claude Code or Codex via pluggable CLI harnesses without rebuilding the underlying runtime.

The AIOS Agent Gateway provides agents with a unified, governed access layer for both models and tools. It connects agents to approved tool environments through the Model Context Protocol (MCP), while routing model requests directly to frontier AI providers such as Anthropic and OpenAI, through enterprise platforms such as Microsoft Foundry and Databricks Unity AI Gateway, or to self-hosted models served using vLLM or Ollama. All connections flow through controlled egress paths to preserve network isolation.

Agent identity and access management

To strengthen security and compliance in banking environments, Zafin AIOS assigns each autonomous agent a distinct, governed digital identity with least-privilege access, auditable activity, and lifecycle controls. This applies workforce-grade identity and access management to multi-agent workflows while reducing the risks associated with shared administrative accounts and generic API keys.

Directory-level identity integration

Every agent in Zafin AIOS is onboarded through the bank’s active corporate directory system (such as Active Directory or LDAP) and possesses a verified email address, system credentials, and a device management profile. This ensures that security teams can provision, audit, or revoke agent permissions using the exact same access-governance tooling used for human staff.

Zero-trust and least-privilege access profiles

Agents inherit default role-based access control guidelines, requiring explicit profile configurations that serve as dynamic boundaries for corporate resource access. These profiles define critical operational parameters, including compute constraints like hardware and network specifications, knowledge access limits for specific document or code bases, and tool entitlements that govern API mutations and database writes. Additionally, they establish workflow permissions to restrict execution paths to authorized tasks, such as bug reproduction or log analysis, ensuring a secure and governed environment.

When an agent changes operational roles, its previous configurations are entirely wiped from the active instance. This ensures that the system enforces least-privilege design rules and prevents privilege escalations during long-running multi-agent processes.

AIOS Agent Knowledge Fabric and drift mitigation

To prevent hallucinations and secure access to the system parameters, policy documentation, code libraries, and database schemas required in complex corporate and investment banking environments, Zafin AIOS utilizes an integrated Knowledge Fabric.

Multi-database grounding layer

The Knowledge Fabric coordinates three distinct database paradigms to support diverse information requests. It utilizes a relational database to store structured transaction schemas, system configurations, document hierarchies, and case history metadata, providing a foundation for organized data management. Complementing this, a vector layer operates semantic embedding indexes to run retrieval-augmented generation pipelines across unstructured corporate policy PDFs, technical runbooks, and historical support records. Finally, a graph system maps systemic relationships, codebase dependency graphs, and regulatory cross-references, ensuring the preservation of structural context across deep operational hierarchies.

Mitigating context poisoning and model drift

A major vulnerability in multi-agent environments is context poisoning, which occurs when incorrect outputs from an active agent are saved back to the database and contaminate future context retrievals. Zafin AIOS mitigates this risk by forcing agents to compile completed tasks into structured case studies.

The platform routes these case studies to an independent LLM-as-a-judge validation engine rather than automatically writing them to the active knowledge base. This engine verifies the document for factual consistency and ensures the resolution aligns with programmatic baseline expectations.

To protect the system from long-term logic regressions, the platform runs scheduled drift detection experiments. These experiments retrieve historical operational scenarios, execute them across multiple local and proprietary model endpoints, and compare the outputs with validated golden-record baselines. This ensures that changes to the knowledge base, model updates, or upstream system modifications do not introduce logical drift or systemic errors.

Connectivity

Connecting autonomous agents directly to enterprise databases, internal transaction ledgers, and third-party SaaS platforms creates complex integration challenges. Ad-hoc API connections demand complex security protocols, specialized error-handling, and unique database adapters.

Zafin AIOS normalizes these systems through a central Tools Registry built around the Model Context Protocol (MCP). MCP functions as an open standard, allowing agents to access diverse software environments through unified server interfaces.

Using MCP simplifies tool integration across the ecosystem. By hosting tools as localized MCP servers, the control plane can enforce standardized auth scopes, error validations, and output protocols. This ensures that agents interact with underlying databases through secure, standardized interfaces, protecting core transactional systems from unguided database mutations.

AIOS Agent Work Management

The execution layer of Zafin AIOS centers on Agent Work Management (AWM), an orchestration engine that manages, schedules, and tracks work items, which are defined as bounded units of work with clear objectives. Rather than treating all tasks identically, the platform organizes work into three distinct operating schedules based on their temporal requirements and scope. Finite work represents standalone processes with defined boundaries, such as modernized code sequences or transaction error triaging, while recurring workflows like weekly vulnerability scans are handled as scheduled cases on a preset calendar. For environmental monitoring and real-time anomaly detection in production logs, the system utilizes continuous work as always-on runtimes, ensuring that every variety of business intent is supported by a specialized execution model.

When an agent becomes idle, it requests work from the AWM engine. AWM recognizes the agent and its associated profile, including its capabilities and the types of work it is authorized to perform. Based on this information, AWM selects and assigns an appropriate work item to the agent.
As the agent executes tasks, the AWM engine tracks state transitions, manages error recovery, and reports active telemetry to the Control Tower. If an agent runs into an operational error or fails to meet a validation gate, the platform routes the task to a coach-and-retry gate, prompting a human operator to step in, provide guidance, and resume execution without restarting the parent work item.

The model gateway

Because frontier model APIs are priced dynamically based on input and output token consumption, running unmanaged multi-agent loops can quickly result in unpredictable computational bills. Zafin AIOS implements an Agent Gateway that acts as a centralized model router and security proxy, decoupling execution requests from specific upstream providers.

The AIOS Agent Gateway enables organizations to use the right model for each task at the appropriate cost point, recognizing that no single model is best suited to every workload. AIOS Evaluations assesses candidate models against task-specific rubrics, and the results inform model selection and configuration. Tasks containing sensitive information are routed to models running on bank-hosted GPUs, while other workloads can use managed or frontier models when their capabilities, performance, and cost are a better fit.

The platform’s Control Tower tracks telemetry for every model call, compiling trace logs that show exact duration metrics, token counts, and costs for every step of a work item. This ensures that bank operators can identify logic loops, locate inefficient model calls, and manage active compute budgets from a single operating dashboard.

Proof of work

Deploying autonomous agents in regulated financial environments requires robust audit processes. Relying on simple post-event text logs is insufficient for compliance validation. Zafin AIOS implements a “proof of work” verification architecture that establishes a secure, chronological, and immutable record of every workflow step.

This verification process ensures that the system records exactly:

1. What was requested and how the platform interpreted the business intent.

2. What specific profiles, models, and tools were active.

3. What context files and database queries were retrieved.

4. What LLM models were used, prompts, inputs, and outputs.

5. What screens looked like during execution via automated video capture.

6. Where and when human reviews occurred in the execution chain.

This comprehensive record travels alongside the work artifacts, giving compliance and audit teams a complete, defensible review trail. This reduces the need for manual post-event audit reconstruction, helping banks meet strict regulatory and compliance guidelines.

Measuring operational performance

To demonstrate the capabilities of the Zafin AIOS platform, the system was validated internally across software development and business operations.

The platform’s performance was tracked across key enterprise metrics.

These results show that establishing a structured control plane turns experimental AI tools into a measurable enterprise operating resource. This architecture enables financial institutions to tackle deferred technical debt, system modernizations, and compliance tasks that were previously too expensive to run with manual engineering resources rather than strictly focusing on head-count reduction.

For banking executives, deploying an end-to-end orchestration platform is a critical infrastructure requirement. While institutions can consume various models from different providers, they must build the internal operating runway and control tower. This ensures that banks can utilize modern AI capabilities in a safe, secure, cost-controlled, and fully governed environment.

Disclaimer:

Fintech Wrap Up aggregates publicly available information for informational purposes only. Portions of the content may be reproduced verbatim from the original source, and full credit is provided with a “Source: [Name]” attribution. All copyrights and trademarks remain the property of their respective owners. Fintech Wrap Up does not guarantee the accuracy, completeness, or reliability of the aggregated content; these are the responsibility of the original source providers. Links to the original sources may not always be included. We use AI tools to brainstorm, draft, edit, and summarize, but the ideas, analysis, and connections Sam draws are his. Every piece is verified and written in Sam’s voice. For questions or concerns, please contact us at sam.boboev@fintechwrapup.com.

In this episode of the WRAP UP podcast, I sit down with Bentzi Rabi, Co-Founder and CEO of Utila, to discuss why institutional wallet infrastructure is becoming one of the most important parts…

Read more

This week’s reports explore how AI agents are reshaping finance and commerce. Building An AI-Ready Finance Function explains that finance teams are moving beyond using AI for automation toward building AI-driven workflows and real-time decision support, though fragmented data and skills gaps remain major barriers. Agentic Payments from the Ground Up examines the rise of machine-to-machine payments, arguing that AI agents and low-cost blockchain infrastructure are making micropayments economically viable for the first time, enabling autonomous software transactions. Agentic Commerce: Who Gets To Transact looks at the emerging infrastructure needed for AI agents to complete purchases independently, highlighting that future success depends not only on payment technology but also on solving challenges around identity, settlement, governance, and liability as autonomous commerce becomes more widespread.

Video of the Week

Deep Dive of the Week

The EU and UK Expense/Spend Management Companies Directory

I built a structural map of the spend management category because the space has quietly split into distinct business models that get lumped together under one label. Some of these companies are regulated e-money institutions issuing their own cards. Others are pure software layers that never touch a BIN. This guide breaks down 22 platforms across five segments, with license type, card networks, card types, features, and regional coverage for each. I also built a companion Excel file with the full dataset so you can filter and compare without scrolling through prose.

This week’s reports

1️⃣Fast Payment Systems as Macroeconomic Infrastructures

2️⃣Agentic AI and the future of B2B payments

3️⃣From AI promise to AI performance

4️⃣Stablecoins: Transforming the Financial Landscape

5️⃣Building An AI-Ready Finance Function

6️⃣Agentic Payments from the Ground Up

7️⃣Agentic Commerce: Who Gets To Transact

Fast Payment Systems as Macroeconomic Infrastructures

Fast payment systems—platforms that settle money transfers within seconds, around the clock—are spreading rapidly across the world. More than 100 countries now operate or are building such systems. Brazil’s Pix, India’s Unified Payments Interface, and similar platforms in the European Union, the United Kingdom, Singapore, and many emerging markets are already handling billions of transactions each year. This paper asks whether that shift in how payments work could also shift how economies perform.

The short answer, as a hypothesis supported by a growing body of indirect evidence, is yes, but only under the right conditions, and the magnitude of the effect is still to be measured. This paper does not claim that fast payment systems have already been proven to raise growth and employment. Its purpose is more modest: to explain why fast payments may matter for the wider economy, through which channels they may operate, and what conditions must be in place for their benefits to materialize. Some of the supporting evidence comes directly from experiences with fast payment systems; some comes from related areas such as digital payments, mobile money, open banking, and fintech credit. A companion empirical paper will test these channels more directly

The Central Hypothesis

Fast payment systems may act as macroeconomic infrastructure—not just a faster way to pay—by simultaneously accelerating the following two fundamental flows that underpin economic activity:

• Money circulation: By settling payments in seconds rather than days, fast payment systems reduce the time that funds sit idle in the payment pipeline. Money becomes usable sooner, reducing the need for households, firms, and banks to keep extra liquidity idle as a precaution.

• Economic information: Real-time, structured transaction records make economic activity more visible and verifiable. Lenders can see a borrower’s actual cash flows, governments can monitor tax compliance, and regulators can spot risks earlier.

These flows are called liquidity acceleration and information acceleration—the two pillars of this paper’s framework.

What the Paper Argues

Read more

I built a structural map of the spend management category because the space has quietly split into distinct business models that get lumped together under one label. Some of these companies are regulated e-money institutions issuing their own cards. Others are pure software layers that never touch a BIN. This guide breaks down 22 platforms across five segments, with license type, card networks, card types, features, and regional coverage for each. I also built a companion Excel file with the full dataset so you can filter and compare without scrolling through prose.

What this guide covers

I looked at 22 companies spanning European and UK SMB card platforms, European spend management tools, enterprise T&E and spend suites, AP automation platforms with card layers, and travel and expense tools. For each one I pulled license type and issuing jurisdiction, card networks supported, card types offered, core feature set, and regional coverage, sourced from official sites and public disclosures. Where a company does not disclose something, I marked it “Not disclosed” rather than guess.

The accompanying Excel file has seven tabs: Overview, Licenses, Card Networks, Card Types, Features, Regions, and a Master Table that puts every company against every dimension in one filterable view.

Subscribe to get access to the directory

The 22 companies

European and UK SMB and mid-market card and spend platforms

Pleo issues Mastercard-branded company cards from its own FCA-authorised e-money entity, aimed at SMB finance teams across Europe. Soldo runs prepaid Mastercard company cards through its own EMI licenses in the UK and Ireland, with a focus on proactive spend control. Moss combines corporate credit cards with AI-driven AP automation for SMEs in Germany, the UK, Austria, and the Netherlands. Mynt issues Visa business cards from its own Swedish e-money license, built around Nordic tax and bookkeeping compliance. Mooncard provides configurable Visa cards for French and European SMEs, issued through Belgian EMI Paynovate. Expend offers a prepaid Mastercard alongside a bring-your-own-card option for UK small businesses. Apron combines invoice capture, bill pay, and an expense card for UK SMEs and their accountants.

European and UK spend management platforms

Spendesk unifies corporate cards, expenses, AP, and procurement for European finance teams, running its own regulated payment institution in France. Payhawk delivers cards, expenses, AP, and procurement to multi-entity mid-market and enterprise businesses, with banking infrastructure via J.P. Morgan. Webexpenses pairs a mature expense-report core with an add-on Mastercard cashback card for mid-market and enterprise teams. N2F runs expense reports, corporate cards, and AP as three modules for companies across 86 countries. Mobilexpense (part of Visma) offers three product tiers spanning SME to large enterprise, with Visa cards issued through Finnish EMI Pliant.

Enterprise T&E and spend platforms

SAP Concur is the dominant enterprise travel, expense, and invoice suite, operating as a software layer over third-party card programs. Navan unifies travel booking with corporate cards and expense automation, issuing through a multi-partner model including Stripe and Adyen. Perk (formerly TravelPerk, after acquiring Yokoy) merges travel booking with AI-driven spend management for scaleups and enterprises. Brex, now a Capital One subsidiary, delivers real-time budgets and AI-reviewed corporate cards through multiple partner-bank issuers.

AP automation and spend platforms

Tipalti automates global mass payouts, AP, procurement, and expenses for mid-market and enterprise finance teams, without issuing its own cards. Stampli layers a card product onto an ERP-aligned procure-to-pay platform built to extend, not replace, a company’s existing system of record.

Travel and expense platforms

Expensify runs a bring-your-own-card model alongside its own Visa Commercial Card, issued through Bancorp Bank and regional partners. ExpenseIn combines expense claims with an integrated spend card, issued through Stripe’s UK and Ireland e-money entities. Zoho Expense is a pure SaaS expense and travel tool with no card issuance, built for breadth across 150+ countries. Airwallex sits partly in this category and partly underneath it, offering multi-currency cards and expense tools while also operating as a licensed BIN sponsor to other platforms.

What’s behind the subscribe button

Subscribers get the full structured profile for each of the 22 companies: what they do, license type and issuing jurisdiction, card networks, card types, key features, and regional coverage. That includes the underlying issuing entities most of these platforms don’t put on their homepage (Stripe, Adyen, Cross River, Sutton Bank, Celtic Bank, Pliant, Paynovate, and others), plus the companion Excel file with all 22 companies mapped across every dimension in a filterable master table

Upgrade to read the full directory of all 22 platforms and download the companion Excel file.

Read more

In this episode of the WRAP UP Podcast, I sit down with Jakob Pethick, Chief Commercial Officer at YouLend, to explore why lending is moving directly into the software businesses already use every day.

Subscribe now

We discuss how platforms such as Amazon, Shopify, eBay, Strip…

Read more

This week’s reports highlight how financial services are entering a new phase of transformation driven by AI, stablecoins, and payment modernization. Banks are investing heavily in fraud prevention, data quality, and instant payment infrastructure, while regulators are developing frameworks to support the safe growth of stablecoins. The reports also explore how digital money is reshaping credit unions, why the UAE is becoming a leader in regulated stablecoin payments, how AI could transform retail financial services by 2030, why payments are becoming a strategic priority for treasury teams, and how robust reconciliation systems are essential for fintechs operating at scale.

Video of the Week

Deep Dive of the Week

Revolut’s Compounding Growth Strategy

Did you read the “Algorithm That Keeps Compounding” case study by Andreessen Horowitz (a16z)? If not, do not worry; I did it for you. In this deep dive, I take analysis by a16z as the core, share my comments on what they got right, what was too optimistic, and what was missing. We have to admit the global consumer fintech market has transitioned from the loose capital era of customer acquisition at all costs to a mature phase defined by structural monetization, regulatory validation, and balance sheet efficiency. While many first generation neobanks struggle under margin compression, the performance metrics of Revolut stands different.

To understand how the compounding engine of Revolut works, we must look past the consumer-facing marketing and dissect the underlying growth algorithm:

This formula is supported by a highly diversified six-segment revenue matrix, an asset-light core architecture, and an operating structure that decouples scale from operational headcount. This deep dive deconstructs these mechanisms, providing a strategic blueprint for fintech founders, product managers, and payments strategists operating in 2026.

This week’s reports

1️⃣2026 KPMG Banking Technology Survey

2️⃣Stablecoins and Macroeconomic Stability

3️⃣How Digital Money Is Impacting Credit Unions

4️⃣The UAE Stablecoin Payments Playbook

5️⃣AI and the future of retail financial services

6️⃣HSBC’s Global Payment Trends Report 2026

7️⃣Account Reconciliation Patterns for High-Volume Fintech

2026 KPMG Banking Technology Survey

KPMG just surveyed 200 banking executives for its 2026 Banking Technology Survey. A few numbers stood out.

91% of banks are prioritizing security and fraud prevention over the next 12 months. Nearly unanimous, and the AI arms race is why. 90% are already using or piloting GenAI for fraud detection, but the same technology is scaling the threat on the other side.

Technology is now an M&A driver, not just a cost center. 33% of banks say tech is directly shaping their acquisition strategy, chasing data/AI capabilities (67%) and cybersecurity (65%).

Data remains the unresolved bottleneck. 100% of respondents cite data quality as a top modernization challenge. You can’t run agentic AI on bad data, and most banks know it.

Payments are moving fast. 72% are modernizing for FedNow, 70% for RTP. Instant payments and ISO 20022 are forcing banks to rebuild rails they’ve patched for decades.

Read more