← Return to the book

FUTURECENTRAL PRESS · BOOK SAMPLE

FinTech in the AI Era

Chapter 17: Banking-as-a-Service and Embedded Finance

How Banking Became a Set of Capabilities Any Platform Can Deliver

Learning Outcomes

By the end of this chapter, the reader will be able to:

Distinguish banking-as-a-service from embedded finance as the supply and the demand sides of one market.

Map the four-party value chain of embedded finance and locate where the customer relationship, the technical ledger, and the regulatory accountability each sit.

Analyze the global banking-as-a-service market, the lesson of the Synapse collapse, and the regulatory tightening that followed.

Trace the Indian banking-as-a-service landscape and the sponsor-bank question that determines which banks participate.

Evaluate the unit economics of banking-as-a-service and judge whether it is a sustainable business.

Opening Vignette

A shopkeeper opens a payment application to check the day's takings and finds, on the same screen, an offer of working capital, preapproved, repayable from the shop's own daily collections. A traveler booking a holiday is shown, at the moment of payment, a travel-insurance policy priced to the exact trip, and accepts it with one tap. A small merchant selling through an online storefront receives an offer of growth capital inside the storefront's own dashboard, underwritten on the sales the platform can already see. None of these three people went to a bank. None of them, in the moment, thought of themselves as dealing with a bank at all.

This is embedded finance, and it represents a quieter and more consequential shift than the neobank story of Chapter 16. The neobank tried to build a better bank that customers would come to. Embedded finance does something more radical: it dissolves the bank into the products people already use, so that the financial service appears at the precise moment of need, inside the application the customer was already in, wearing the brand of a company that is not a bank.

Behind every one of those embedded products, however, a bank is still present, and so is a layer of technology that connects the two. The shopkeeper's loan sits on some licensed lender's balance sheet. The traveler's policy is underwritten by a licensed insurer. The merchant's capital is funded by a regulated institution. What has changed is not that the bank disappeared but that the bank became invisible, reached through an interface, summoned by a platform, and presented to the customer under someone else's name. The machinery that makes this possible is banking-as-a-service, and it is the subject of this chapter.

This chapter is the architectural heart of Part 4. Chapter 16 took up the neobank, digital banking’s visible face. This chapter examines the deeper structure: banking-as-a-service, the supply of banking capabilities as software, and embedded finance, the consumption of those capabilities by the nonfinancial platforms that now distribute financial products. The shift it describes may, in the long run, matter more than the neobank ever did.

This chapter distinguishes banking-as-a-service from embedded finance, maps the four-party value chain that connects them, examines the global market and the Synapse collapse that reshaped it, and assesses the Indian landscape, the sponsor-bank question, and whether banking-as-a-service is a sustainable business.

17.1 Banking-as-a-Service and Embedded Finance

Banking-as-a-service and embedded finance are the two sides of a single market, and a reader must hold the distinction precisely: banking-as-a-service is the supply side, and embedded finance is the demand side. The two terms are often used loosely and sometimes interchangeably, and the looseness obscures the structure. The structure becomes clear the moment the supply and the demand are named separately.

Banking-as-a-service, on the supply side, is the packaging of banking capabilities into software that other firms can consume. A licensed bank holds capabilities that are valuable and hard to replicate: the ability to hold a deposit, to issue a card, to make a payment, to originate a loan, each performed inside a regulatory perimeter. Banking-as-a-service takes those capabilities and exposes them as application programming interfaces (APIs), modular services that a developer at another company can call. The supplier may be a bank offering its own capabilities directly, or, more commonly, a specialist technology firm that packages a bank's capabilities into a developer-friendly form. Either way, what is being supplied is banking, sold as a service rather than as a destination.

Embedded finance, on the demand side, is the consumption of those capabilities by nonfinancial platforms that build financial products into their own customer experience. A retailer, a logistics platform, a travel application, a software tool for small businesses, each has customers, a relationship, and a context, and each can now add a financial product, a payment, a loan, an insurance policy, a card, without becoming a bank. Embedded finance is what the demand side does with what banking-as-a-service supplies, and the embedding platform's motive is examined in the next section.

The term banking-as-a-service describes an idea older than the name, and a sentence of history clarifies it. Banks have always sold capabilities to other firms: a small bank issuing cards under a larger institution's brand, a bank providing the rails behind a corporation's payments. The practice of one regulated institution lending its license and its plumbing to another firm's product is decades old. What is new is the scale, the automation, and the audience. The API turned a capability that was once negotiated firm by firm, slowly and at a large minimum size, into something a developer at a small startup could integrate in days. Banking-as-a-service is the old practice of capability-sharing, industrialized by software and opened to firms that could never before have afforded it.

The two sides meet in a value chain with four parties, and naming the four now sets up the framework the chapter builds. At one end is the sponsor bank, the licensed institution that holds the regulated capability. Next is the banking-as-a-service platform, the technology firm that packages the capability into consumable software. Then comes the embedding platform, the nonfinancial company that presents the financial product to its own customers. At the far end is the end customer, who experiences a financial service inside a nonfinancial product and may never know how many parties stood behind it. The whole of this chapter is, in one sense, an examination of that four-party chain: how value flows along it, how it can fail, and who is accountable when it does.

17.2 What Embedded Finance Actually Is

Embedded finance is best understood through the concrete examples that already exist, because the abstraction conceals how ordinary, and how widespread, the shift has already become. A reader who treats embedded finance as a futuristic concept has not noticed how much of it is already in daily use.

The clearest examples are in lending. A small-business commerce platform that can see a merchant's sales can offer that merchant a loan, underwritten on the sales data the platform already holds, repaid automatically from future sales. Shopify's capital offering and Amazon's lending to its sellers are the global archetypes, and the embedded lending of the Indian PayTech super-apps, examined in Chapter 8, is the same idea: credit offered inside the platform where the borrower already transacts. Embedded lending works because the platform knows the borrower's business better than any outside lender could, and reaches the borrower at the exact moment a loan is useful.

Embedded payments and embedded insurance are equally established. A ride-hailing application, a food-delivery platform, or a hospitality service processes payments inside its own experience, so smoothly that the customer barely registers a payment occurring. Embedded insurance places a policy at the point of sale or the point of booking, a device protection plan as a phone is bought, a travel policy as a trip is booked, priced to the specific purchase and offered at the moment the risk becomes salient. Embedded cards let a platform issue its own branded card, carrying its rewards and its relationship, on rails it does not own.

The embedded-finance thesis is sometimes summarized in a slogan, that every company will become a financial company, and the slogan is worth examining rather than repeating. The formulation traces to Angela Strange of Andreessen Horowitz, who argued in 2019 that every company would become a fintech company. The slogan overstates and understates at once. It overstates, because most companies will not become financial companies in any meaningful sense: they will not hold a license, carry credit risk, or run a treasury. It understates, because the change it points at is real: a great many nonfinancial companies will distribute financial products, earn revenue from them, and weave them so deeply into their offering that the customer experiences the finance and the core product as one thing. The accurate version of the slogan is narrower and more useful. Every company with customers and context will have the option to distribute finance, and the companies whose customers carry a recurring financial need will take it.

The strategic logic for the embedding platform rests on three advantages, and the three together explain why embedded finance is spreading so fast. The first is context: the platform knows what the customer is doing and therefore what financial product is relevant, at a moment a bank could never identify. The second is margin: financial products carry attractive economics, and a platform that adds them monetizes a customer base it has already acquired and paid for. The third is engagement: a financial product woven into a platform deepens the customer's reliance on it and raises the cost of leaving. The embedding platform gains revenue, retention, and relevance, and it gains all three without taking on a banking license.

The shift embedded finance represents can be stated in one sentence, and it is the sentence the chapter asks the reader to carry forward. Finance is moving from a destination to a feature. For most of banking's history, a financial product was something a customer traveled to obtain, to a branch, then to a bank's own website, then to a bank's own application. Embedded finance makes the financial product appear inside whatever application the customer is already using, summoned by the platform that owns the customer's attention. The best distribution for a financial product, in the embedded-finance era, is the platform where the customer already is, and that simple fact is what makes banking-as-a-service a structural shift rather than a passing fashion.

17.3 The Global Banking-as-a-Service Market

The global banking-as-a-service market matured earlier than India's, and its history carries a warning that the entire chapter is built around: the 2024 collapse of Synapse. A reader cannot understand the risks of banking-as-a-service, or the regulatory tightening now reshaping it, without understanding what happened when a banking-as-a-service firm failed.

The global market produced a set of well-known banking-as-a-service firms. In the United States, Synapse, Treasury Prime, and Galileo, the last now owned by the fintech SoFi, built businesses packaging bank capabilities for fintech developers. In Europe, Solaris became the largest banking-as-a-service provider, and card-issuing specialists such as Marqeta built substantial businesses on a single capability. For a period, banking-as-a-service was one of the most enthusiastically funded categories in fintech, on the thesis that every company would, in time, become a financial company.

The card-issuing specialists deserve a separate mention, because they show banking-as-a-service working in a narrower and safer form. Alongside the broad banking-as-a-service middleware firms, a set of specialists built businesses around a single capability, the issuing of cards, and Marqeta is the best-known of them. A card-issuing platform is a more contained proposition than a full banking-as-a-service stack: it does one well-defined thing, its technical and regulatory surface is narrower, and the reconciliation problem that destroyed Synapse is correspondingly smaller. The relative stability of the card specialists, set against the broad middleware firms' troubles, is an early hint of a lesson the chapter's framework will sharpen: the danger in banking-as-a-service rises with the number of distinct capabilities, and the number of distinct parties, that a single arrangement tries to hold together.

The Synapse collapse of 2024 was the event that ended the enthusiasm and exposed the model's central risk. Synapse was a banking-as-a-service middleware firm, the technology layer between consumer fintech applications and the banks that actually held the deposits. When Synapse failed and entered bankruptcy, a catastrophic problem emerged: the records could not be reconciled. The ledgers maintained by the fintech applications and the funds actually held at the partner banks did not match, and tens of millions of dollars of ordinary people's deposits could not be cleanly accounted for.1 Customers of the affected applications were locked out of their own money for months. The failure was not that a technology firm went bankrupt, which firms do. The failure was that when the middleware vanished, no one could say with certainty who owned which dollar.

The Synapse collapse revealed the structural vulnerability of the banking-as-a-service model, and the chapter states it plainly because the framework later formalizes it. In a banking-as-a-service arrangement, the customer-facing brand, the regulated bank, and the firm that keeps the connecting ledger can be three different parties. When they are, the ledger that reconciles the customer to their money sits with the middle party, and if the middle party fails, the customer's brand cannot help them and the bank may hold the deposits without holding the records that assign them. Synapse was that failure made real, and it is the reason banking-as-a-service is now among the most scrutinized activities in finance.

Solaris, in Europe, told a quieter version of the same cautionary story. The largest European banking-as-a-service provider faced sustained regulatory pressure and operational difficulty, and required a strategic reset and fresh capital raised under strain. Solaris did not collapse as Synapse did, but its difficulties confirmed that banking-as-a-service is a demanding business, heavy in compliance, thin in margin, and unforgiving of weak controls. The global market's lesson, carried into the rest of the chapter, is that banking-as-a-service is genuinely useful infrastructure and genuinely dangerous when its controls and its accountability are not exact.

17.4 The Indian Banking-as-a-Service Landscape

India's banking-as-a-service landscape is led by specialist technology firms rather than by the banks themselves, and a small set of those firms supplies most of the capability that Indian embedded finance runs on. Where the global market produced banking-as-a-service firms that often sat close to a single sponsor bank, India produced a layer of independent infrastructure companies that connect many fintechs to many banks, and three names define it.

M2P Fintech is the scaled leader of Indian banking-as-a-service, and it anchors this chapter. M2P built its business on issuing-and-acquiring-as-a-service: the technology to issue cards and accounts and to process payments, offered as software that banks, lenders, and fintechs can consume. By the mid-2020s its infrastructure powered more than 200 banks, more than 300 lenders, and several hundred further fintech engagements, serving end users numbering in the tens of millions.2 M2P is the Indian banking-as-a-service platform at genuine scale, and the chapter's anchor case examines it in full.

Setu and Decentro are the other significant Indian banking-as-a-service firms, and each took a distinct path. Setu built APIs for the India Stack and for banking functions, payments, deposits, data-sharing, and was subsequently acquired by the larger payments firm Pine Labs, an instance of the consolidation that recurs across this book's infrastructure chapters. Decentro built a developer-first banking-and-payments API platform, competing on the speed and the ease with which a developer can integrate. Cashfree and other payments firms have added banking-as-a-service capabilities alongside their core businesses. The Indian landscape is, in short, a layer of independent API firms, consolidating gradually, that sit between the country's banks and its fintechs.

India's banking-as-a-service layer was shaped, more than any other country's, by the India Stack. The public digital infrastructure examined in this book's early chapters, the identity layer, the Unified Payments Interface (UPI), the Account Aggregator framework, gave Indian banking-as-a-service firms a foundation no other market handed their equivalents. An Indian banking-as-a-service platform did not have to build identity verification, a real-time payment rail, or a consented data layer for itself; it could build on the public rails and concentrate its effort on the packaging and the integration that are its actual value. The India Stack is part of why India produced a layer of independent banking-as-a-service firms rather than firms each tethered to a single bank: the hardest shared plumbing was already public, and the banking-as-a-service firms competed on what they added above it.

The Indian model rests on the sponsor-bank architecture, and the regulator holds the sponsor bank accountable for it. An Indian banking-as-a-service arrangement requires a licensed bank behind it, because the regulated activity, holding the deposit, issuing the instrument, can be performed only by a licensed entity. The banking-as-a-service firm supplies the technology; the sponsor bank supplies the license and carries the regulatory accountability. This is the same accountability principle the book has met in the LSP-RE architecture of digital lending: the Reserve Bank of India cannot license every fintech, so it holds the licensed bank answerable for the fintechs it enables.

The Reserve Bank's posture toward banking-as-a-service has tightened, and the tightening is the most important recent development in the Indian landscape. Through 2024, the Reserve Bank examined bank-fintech partnership arrangements closely and acted against specific arrangements it judged poorly governed, including in the area of co-branded cards and prepaid programs.3 The regulator's concern is precisely the concern the Synapse collapse made vivid: an arrangement in which the customer-facing fintech, the technology layer, and the licensed bank are loosely coordinated is an arrangement in which accountability can blur and control can fail. The Indian regulator, watching the global experience, has chosen to discipline the model before a domestic Synapse occurs rather than after, and the framework that follows explains exactly what it is disciplining.

Framework: The Four-Party Embedded Finance Stack

The Four-Party Embedded Finance Stack is the framework for analyzing any embedded-finance arrangement, and its central purpose is to locate three things that, dangerously, can sit with three different parties: the customer relationship, the technical ledger, and the regulatory accountability. The chapter has named the four parties. The framework now sets out what each contributes, what each earns, and, most importantly, where the points of failure lie.

The first party is the sponsor bank, and it contributes the license and the balance sheet. The sponsor bank is the only party in the stack that can lawfully perform the regulated activity, hold the deposit, issue the card, book the loan, and it is the party the regulator holds accountable for the whole arrangement. It earns a fee or a share of the economics. Crucially, the sponsor bank is the party the end customer never sees, and that invisibility is the first source of danger: the accountable party is the hidden one.

The second party is the banking-as-a-service platform, and it contributes the technology and, often, the ledger. The banking-as-a-service platform packages the bank's capability into consumable software and operates the technical layer that connects the embedding platform to the bank. In many arrangements it also keeps, or helps keep, the ledger that records which end customer owns which balance. It earns a platform fee and a transaction fee. The banking-as-a-service platform sits in the middle of the stack, and the Synapse collapse showed what happens when the middle fails.

The third party is the embedding platform, and it contributes the customers and the context. The embedding platform is the nonfinancial company, the retailer, the marketplace, the software tool, that presents the financial product to its users inside its own experience and under its own brand. It earns a margin or a commission on the embedded product. The embedding platform is the party the customer believes they are dealing with, which is the second source of danger: the visible party is not the accountable one.

The fourth party is the end customer, who experiences a single seamless financial product and is, by design, shielded from the complexity behind it. The customer's experience is the measure of the arrangement's success and, when the arrangement fails, the customer is the party with the least power to put it right, because the customer's contract is with a brand that holds no license, the ledger is with a firm the customer has never heard of, and the deposits are with a bank the customer did not know existed.

A worked example shows the framework locating a risk before it becomes a failure. Consider a popular savings application offered by a well-known consumer brand. The customer believes the savings are with the brand, because the brand is all the customer sees. The deposits are in fact at a sponsor bank the customer has never heard of. The ledger that records which customer owns which rupee is maintained by a banking-as-a-service platform the customer has also never heard of. Apply the framework's three locations: the relationship sits with the brand, the accountability sits with the bank, and the ledger sits with the middleware. Now apply the diagnostic question. If the middleware firm failed tomorrow, could the sponsor bank, from its own records, tell each customer what they own? If the bank holds only a single pooled balance and relies entirely on the middleware's ledger to divide it, the answer is no, and the framework has identified a Synapse risk inside a product that, to every customer using it, looks perfectly safe.

The framework's central discipline follows from that separation, and it is a single diagnostic question. In a sound embedded-finance arrangement, the roles are unambiguous and, above all, the ledger that reconciles each customer to their money can be reconstructed by the sponsor bank independently of the banking-as-a-service platform. The diagnostic question to ask of any arrangement is therefore this: if the banking-as-a-service platform vanished overnight, could the sponsor bank and the customer still establish, exactly, who owns what? Where the answer is yes, the arrangement is sound. Where the answer is no, the arrangement is a Synapse waiting to happen, and the framework's verdict is that no embedded-finance product should be launched until the answer is yes. The separation of relationship, ledger, and accountability is what makes embedded finance powerful, and disciplining that separation is what keeps it safe.

17.5 The Sponsor Bank Question

Not every bank chose to become a sponsor bank for embedded finance, and the question of which banks did, and why, reveals the strategic logic of the sponsor role. A reader who assumed that every bank would rush to supply banking-as-a-service would be wrong, and the pattern of who participated is informative.

The banks that became banking-as-a-service hubs in India were, characteristically, the mid-sized and smaller ones. Banks such as RBL Bank, Equitas Small Finance Bank, Federal Bank, and others positioned themselves as the licensed partners behind fintech arrangements, and the reason is a matter of strategic fit. A mid-sized bank lacks the vast branch network and the powerful brand of the largest banks, and so it lacks their ability to acquire customers, deposits, and fee income directly. Banking-as-a-service offers such a bank a different route to all three: it can reach customers, gather deposits, and earn fee income through the distribution of dozens of fintech partners, distribution it could never have built itself. For the mid-sized bank, embedded finance is a growth strategy.

The largest Indian banks have been far more cautious about the sponsor role, and their caution is equally rational. A bank that already has enormous scale, a trusted brand, and a national branch network gains less from fintech-partner distribution, because it already reaches the customers. Against that modest gain it must weigh a real cost: the regulatory and reputational risk of being accountable for the conduct of many fintech partners, the risk the Reserve Bank's 2024 scrutiny made concrete. For a large bank, the arithmetic frequently says that the sponsor role is not worth the exposure.

Becoming a sponsor bank, moreover, requires capabilities that not every bank possesses. A bank that wishes to supply banking-as-a-service at scale needs modern APIs that fintechs can integrate cleanly, a technology estate that can support many partners at once, and a risk-and-compliance function genuinely capable of governing those partners rather than merely contracting with them. A bank whose systems are old and whose risk function is built only for its own direct business cannot safely be a sponsor bank, however much it might want the fee income.

The sponsor-bank role also changes a bank's risk profile in a way that deserves to be made explicit. A bank that lends and takes deposits directly carries risks it can see and price: its own borrowers, its own customers. A sponsor bank carries, in addition, the conduct risk of every fintech partner it enables, and that risk is harder to see, because the partner, and not the bank, stands in front of the customer. If a fintech partner mis-sells, mishandles data, or collects abusively, the regulatory and reputational consequence reaches the sponsor bank, which may learn of the conduct only after it has occurred. The sponsor role converts a portfolio of fintech contracts into a portfolio of conduct exposures, and a bank that has not built the capability to supervise that portfolio has accepted a risk it cannot observe in time to manage.

The sponsor-bank question carries a warning that the chapter's regulatory section develops. The sponsor bank earns its fee in exchange for carrying the accountability of the whole stack, and a bank that pursued the fee without building the capability to govern its partners took on an exposure it could not control. The Reserve Bank's actions of 2024 fell, in part, on exactly such banks. The lesson for a bank weighing the sponsor role is that banking-as-a-service is a real and legitimate business, and it is a business that punishes a sponsor bank that treats partner governance as a formality rather than as the core of the role.

17.6 Unit Economics and the Banking-as-a-Service Profitability Question

The unit economics of banking-as-a-service are demanding, and the chapter must ask directly whether banking-as-a-service is a sustainable business or a structurally difficult one. The category attracted enormous investment on the thesis that every company would become a financial company, and the thesis is broadly sound. Whether the firms that supply the capability can themselves prosper is a separate question, and a harder one.

The banking-as-a-service platform earns a thin take rate, and thinness is the defining fact of its economics. The platform charges a fee on each transaction, each account, or each card, and a platform fee for access, and each individual fee is small. A banking-as-a-service business is therefore a volume business in the most demanding sense: it becomes profitable only at very large scale, because only very large transaction volume turns a thin per-unit fee into a meaningful total. A banking-as-a-service firm below the necessary scale is a firm losing money in every reporting period, funding the gap with investor capital while it races toward a volume it may not reach.

The cost side of the banking-as-a-service business is heavy, and it is heavy in ways that do not shrink with scale. A banking-as-a-service platform must build and maintain the technical integration to many banks and many fintechs, and it must run a compliance and risk-management operation substantial enough to satisfy the regulator and the sponsor banks. After the Synapse collapse, the compliance burden rose sharply across the industry, because every party now demands stronger controls. The combination, a thin take rate against a heavy and rising cost base, is what makes banking-as-a-service a structurally hard business.

The build-versus-buy decision of the embedding platform also bears on the banking-as-a-service firm's economics, and it cuts both ways. A nonfinancial platform deciding to add embedded finance can integrate a banking-as-a-service provider's software, which is fast and avoids the platform becoming a regulated entity, or it can build deeper financial capability itself, which is slow and costly. For most embedding platforms the buy decision is correct, which is the banking-as-a-service firm's opportunity. But the largest embedding platforms, at sufficient scale, may conclude that building, or acquiring a banking-as-a-service capability outright, is cheaper than paying a per-transaction fee forever, which is the banking-as-a-service firm's standing threat, the same threat Chapter 14 identified for LendTech infrastructure.

The investor's view of banking-as-a-service shifted sharply after Synapse, and the shift is part of the economic picture. Through the early 2020s, banking-as-a-service was funded generously, on the thesis that the embedded-finance wave would lift every supplier of the capability. The Synapse collapse changed the calculation. Investors learned that the most serious risk a banking-as-a-service firm carries is a control failure that can destroy the firm and harm thousands of customers overnight, a risk graver than slow growth, and they began to price it. Capital became more selective, valuations compressed, and the firms that could raise on good terms were those that could demonstrate scale, profitability, and, above all, the integrity of their reconciliation. The post-Synapse funding environment is itself a discipline on the category, rewarding the careful and starving the careless, and it is part of why the chapter expects the survivors to be few.

The honest answer to the profitability question is that banking-as-a-service is a genuine and necessary layer that is nonetheless a hard business, and that the survivors will be few. The Synapse collapse and the Solaris difficulties show what happens to banking-as-a-service firms that cannot carry the cost base or control the risk. The scale and market leadership that M2P has built, examined in the anchor case, show that the business can be carried toward viability, but only with scale, with disciplined risk management, and with the diversification that protects a firm against the loss of any single large customer. Banking-as-a-service is infrastructure the embedded-finance era genuinely requires, and it is infrastructure that only a disciplined and well-scaled firm should attempt to supply.

17.7 The Three Waves in Banking-as-a-Service

The three waves of AI run through banking-as-a-service along two arcs at once: the risk controls that keep the model safe, and the developer experience that determines how fast it grows. A banking-as-a-service platform is, beneath the regulatory packaging, a software business serving developers and a risk business serving banks, and AI has advanced both.

Wave 1 in banking-as-a-service was manual integration and rule-based risk. In the first wave, connecting a fintech to a bank through a banking-as-a-service platform was a slow, hand-built engineering project, and the risk screening applied to the transactions that flowed through the platform was rule-based, catching the patterns the rules anticipated and missing the rest. The model worked, and it worked at the pace and the accuracy that manual engineering and fixed rules allowed.

Wave 2 introduced machine learning, and it strengthened both arcs. On the risk arc, machine-learning models screen the transactions flowing through the banking-as-a-service platform for fraud and for the signatures of misuse, and they assess the credit risk of embedded lending with the methods Chapter 11 described. On the integration arc, machine learning helps route and configure connections more intelligently. Wave 2 made the banking-as-a-service platform both safer, because its risk screening improved, and faster to deploy, because its integration became less purely manual.

Generative AI defines Wave 3, acting on the developer experience and the risk narrative. A generative system can assist a developer integrating a banking-as-a-service platform by generating integration code, answering questions in natural language, and diagnosing errors, compressing an integration that once took weeks. On the risk arc, generative AI can synthesize the narrative explanations that compliance and supervision require, turning the raw output of a risk model into the documented reasoning a sponsor bank and a regulator expect. Wave 3 lowers the cost of the two things a banking-as-a-service platform spends most on: bringing a new partner live, and satisfying the compliance burden that the post-Synapse era has raised.

A closing observation ties the three waves back to the chapter’s economics. Section 17.6 established that banking-as-a-service is a hard business, squeezed between a thin take rate and a heavy cost base. The three waves act directly on that cost base: AI reduces the cost of integration and the cost of compliance, the two largest costs a banking-as-a-service platform carries. For a category whose profitability depends on scale, a technology that lowers the cost of adding each new partner and each new check goes to the viability of the business itself.

17.8 The Agentic Frontier

The agentic frontier in embedded finance is the prospect of AI agents transacting across embedded financial services, and it adds a fifth actor to the four-party stack the chapter has built. Chapter 9 examined the agent as a payer; here the agent appears instead as a consumer of embedded finance, and that shift reshapes the framework.

An agentic embedded-finance transaction is straightforward to picture. An agent acting for a customer transacts on an embedding platform, and in doing so it consumes the platform's embedded payment to settle, perhaps its embedded credit to fund the purchase, perhaps its embedded insurance to protect it. The financial products woven into the platform are now being summoned by software rather than chosen by a person, and the agent moves through the embedded-finance stack at machine speed, across many platforms, without a human pausing at any step.

The agent becomes a fifth party in the stack, and its arrival sharpens the framework's central problem. The Four-Party Stack already separated the customer relationship, the technical ledger, and the regulatory accountability across different parties. The agent inserts a further layer between the end customer and the embedding platform, and it raises every question the framework asked. Whose instruction did the agent act on, and how is that authority proven? When an agent triggers embedded credit, who has assessed that the customer can repay? When something goes wrong, the customer's recourse must now run back through the agent, then the embedding platform, then the banking-as-a-service platform, then the sponsor bank, a longer chain than the one the Synapse customers could not navigate.

The chapter's judgment is that agentic AI deepens the embedded-finance thesis and sharpens its accountability problem at the same time, and both must be held together. Agentic transactions will increase the volume flowing through embedded finance, which is good for a banking-as-a-service platform whose economics depend on volume. They will also stretch the four-party stack into a five-party one, and the diagnostic question of the framework, whether the chain can be reconciled if the middle fails, becomes harder to answer with every layer added. The agentic era will make embedded finance larger and will make the discipline of the Four-Party Stack more necessary, not less. A banking-as-a-service platform that grows its agentic volume without tightening its reconciliation has simply built a bigger version of the structure that failed in 2024.

Anchor Case: M2P Fintech

M2P Fintech is the right anchor for this chapter because it is the largest Indian banking-as-a-service platform, it has reached genuine scale and market leadership in a category that has destroyed firms, and its international expansion tests whether Indian banking-as-a-service architecture can travel. A reader who follows M2P from its founding to its present scale sees the banking-as-a-service thesis demonstrated, and demonstrated by a firm that has so far avoided the fate of Synapse and the difficulties of Solaris.

M2P was built around a clear thesis: that issuing and acquiring should be available as a service. Founded by Madhusudanan R and his cofounders, M2P set out to supply, as software, the capabilities a firm needs to issue a card or an account and to process a payment. A bank, a lender, or a fintech that wanted to launch a card program, a wallet, or a payment flow could, instead of building the capability over years, integrate M2P's platform and launch in a fraction of the time. This is banking-as-a-service in the precise sense of Section 17.1: a hard banking capability, packaged into consumable software.

M2P's scale, by the mid-2020s, made it the reference point for Indian banking-as-a-service. Its infrastructure powered more than 200 banks, more than 300 lenders, and several hundred further fintech engagements, and it served end users numbering in the tens of millions.2 In the framework's terms, M2P is the banking-as-a-service platform party in a very large number of four-party stacks at once, the technology layer connecting India's banks to India's fintechs. Few firms sit behind as much of the country's embedded finance.

The firm's scale and growth are what separate it from the cautionary cases, and they are the reason the chapter can anchor on it rather than merely cite it. M2P reported revenue of approximately ₹382 crore for the financial year ending March 2024, a decline of roughly 13 percent from the prior year, before revenue recovered to more than ₹500 crore in the year ending March 2025. Like the banking-as-a-service category globally, the firm remained loss-making as it invested in growth and international expansion.4 Section 17.6 explained why banking-as-a-service is a hard business, thin in margin and heavy in cost. M2P is the evidence that the business can be built to genuine scale and market leadership, on a customer base diversified too broadly for any single loss to threaten, which is the precondition of eventual viability even where viability has not yet been reached.

M2P's funding history shows a firm being capitalized for international expansion rather than for survival. M2P raised a Series D round of roughly ₹850 crore, on the order of 100 million dollars, announced in September 2024 and led by Helios Investment Partners, at a valuation that surpassed ₹6,550 crore, and it has raised on the order of 200 million dollars in total across its history.5 The identity of the lead investor is itself informative: Helios is an Africa-focused investor, and the round was explicitly directed at M2P's expansion into Africa's underdeveloped fintech market, alongside its existing growth in the Middle East and its ambitions in Southeast Asia.

The firm's growth has been partly a story of acquisition, and the pattern matters. M2P built its breadth not only by developing capabilities but by acquiring firms across the adjacent parts of the stack, assembling a fuller platform than any single product line would have produced. The strategy is the one Chapter 14 identified at Perfios: in an infrastructure category that rewards scale and breadth, the firm that acquires its way to a full stack offers its customers one integration and one accountable counterparty in place of many. M2P's acquisitiveness is therefore the logical execution of the banking-as-a-service economics rather than mere opportunism, because a broad, scaled platform is worth more, to a customer and to an investor, than a strong but narrow one.

M2P's international expansion is the strategic test the chapter cares about most, because it asks whether Indian banking-as-a-service architecture can export. By 2025, M2P's leadership stated that international markets contributed roughly 40 percent of revenue, with the firm's own ambition pointing toward a substantially higher share.4 This matters because it answers a question the book has raised about Indian fintech repeatedly. A customer-facing Indian fintech is hard to export, because it is bound to one market's customers and regulation. M2P is testing whether an Indian banking-as-a-service platform, selling a capability rather than a consumer product, can carry its architecture into the Middle East, Africa, and Southeast Asia. Its progress so far suggests that infrastructure travels where customer-facing fintech does not, which is the same lesson Chapter 14 drew from Perfios.

M2P's strategic positioning, set against Setu and Decentro, completes the case. Setu built a focused API business and was absorbed by a larger payments firm; Decentro competes on developer experience; M2P pursued scale and breadth and remained independent. The contrast is the chapter's argument about banking-as-a-service economics made concrete: in a business that rewards scale and punishes the subscale, the firm that reached genuine scale and genuine diversification is the firm still standing and still independent. M2P is the anchor case because it shows that the hard business of banking-as-a-service can be built to real scale and market leadership, and that, done with discipline, it can be exported, even though the category, M2P included, has yet to prove it can be reliably profitable.

India Cases

Three further Indian cases complete the banking-as-a-service landscape, two of them banking-as-a-service firms and one a sponsor bank, because the stack cannot be understood from the technology layer alone. Together with M2P they map the Indian model.

Setu shows the focused API specialist, and its absorption shows the category consolidating. Setu built clean, well-regarded APIs for banking and for the India Stack, payments, deposits, data-sharing, and it built a reputation among developers for the quality of its engineering. It did not, however, pursue the scale-and-breadth strategy that M2P did, and in time it was acquired by Pine Labs, the larger payments and merchant-services firm.6 Setu's path is the path Chapter 14 identified for infrastructure specialists generally: a strong firm in a narrow part of the stack, in a consolidating category, becomes a component of a larger platform. The quality of Setu's engineering was real, and it was not, on its own, a substitute for scale.

Decentro shows the developer-experience strategy, the bet that the easiest platform to integrate wins. Decentro built a banking-and-payments API platform with a deliberate focus on the speed and the simplicity with which a developer can integrate it, on the thesis that, among banking-as-a-service platforms offering broadly similar capabilities, the one a developer can adopt fastest will win the developer's business. It is a coherent strategy in a category where integration cost is a genuine burden, and it positions Decentro as the developer-first alternative to the scaled breadth of M2P. Whether developer experience alone can carry a firm to the scale that Section 17.6 showed banking-as-a-service requires is the open question of the Decentro case.

RBL Bank shows the sponsor-bank side of the stack, and the risks of the role. RBL Bank was among the Indian banks that embraced the sponsor role, building a substantial business behind fintech partnerships, notably in co-branded credit cards. That business gave RBL reach and fee income it could not have built through its own branch network, the strategic logic of Section 17.5. It also exposed the bank to the governance burden of the sponsor role, and co-branded card arrangements across the industry drew regulatory attention during the Reserve Bank's broader scrutiny of bank-fintech partnerships.7 RBL's experience is the sponsor-bank question made concrete: the role offers a mid-sized bank genuine growth, and it offers that growth only to a bank prepared to govern its partners as rigorously as it governs itself. The sponsor bank earns its fee by carrying accountability, and a bank that takes the fee without building the governance has taken a risk the regulator will eventually price.

Global Cases

The global cases of this chapter divide into a warning and an opportunity: the failures that taught the industry what banking-as-a-service can do wrong, and the embedding platforms that show what embedded finance can do at scale. Both halves are necessary, because the chapter has argued that banking-as-a-service is genuine infrastructure and a dangerous business at once.

Synapse is the warning, and it deserves a fuller account than the chapter has yet given. Synapse Financial Technologies was a banking-as-a-service middleware firm in the United States, the layer between consumer fintech applications and the banks that held the deposits. When Synapse entered Chapter 11 bankruptcy in April 2024, the failure that mattered was what the bankruptcy exposed rather than the bankruptcy itself: a reconciliation gap, in which the balances recorded by the fintech applications and the funds actually held at the partner banks could not be matched.8 The bankruptcy trustee estimated the shortfall at between $65 million and $96 million; ordinary customers across several fintech applications were frozen out of their savings for months, and many recovered only a portion of what they had believed was theirs. Synapse is the case the Four-Party Stack framework was built to explain: the customer-facing brand, the ledger-keeping middleware, and the deposit-holding bank were three different parties, and when the middle one failed, the chain could not be reconstructed.

Solaris is the quieter warning, the largest European banking-as-a-service firm meeting the limits of the model. Solaris did not collapse, but it spent the mid-2020s under sustained regulatory pressure from its supervisor, working through operational difficulties, and raising capital under strain to fund a strategic reset that ended its independence: a €140 million financing round in early 2025 made Japan's SBI Group the majority owner.9 Solaris confirms what Section 17.6 argued: banking-as-a-service is a business heavy in compliance and thin in margin, and even the category's European leader found that combination punishing.

It is worth drawing the precise lesson the two failures share, because they failed in different ways for the same underlying reason. Synapse failed suddenly and Solaris struggled slowly, but both difficulties trace to a single source: a banking-as-a-service business carries an obligation, to keep its records, its controls, and its accountability exact, that does not scale down. A small banking-as-a-service firm has the same reconciliation duty and the same regulatory exposure as a large one, with a fraction of the resources to discharge them. The category therefore punishes the subscale and the undercontrolled with particular severity, and the two global failures are, between them, the clearest statement of why Section 17.6 concluded that banking-as-a-service is a business in which only the scaled and the disciplined survive.

Shopify Capital and Amazon Lending are the opportunity, the embedding platforms that show embedded finance at full scale. Shopify offers its merchants capital, underwritten on the sales data Shopify already holds and repaid from future sales, and Amazon does the same for its sellers.10 Neither Shopify nor Amazon is a bank; each works with licensed lending partners for the capital while supplying the customer relationship, the context, and the data. They are the embedding platforms of the framework, operating at a scale that makes them among the most significant lenders to small businesses in their markets. The two cases together state the chapter's balance: embedded finance is a genuinely powerful distribution model for the platform that owns the customer, and banking-as-a-service is the genuinely hard business of supplying the capability behind it.

Regulatory Landscape

The regulation of banking-as-a-service has moved, since the Synapse collapse, from a peripheral concern to a global priority, and the four jurisdictions show a clear convergence toward holding the licensed bank firmly accountable for the whole stack. That convergence is what the two-tier comparison below is built to trace.

The Global Layer

The United States tightened sharply after Synapse, and its regulators acted on exactly the vulnerability the collapse exposed. Federal banking regulators intensified their scrutiny of bank-fintech partnerships, brought enforcement actions against partner banks whose oversight of their fintech arrangements had been inadequate, and moved toward stronger recordkeeping requirements for the custodial deposits that sit behind banking-as-a-service, so that the reconciliation failure of Synapse would be harder to repeat.11 The American response is the clearest regulatory statement that the sponsor bank owns the accountability for its fintech partners.

The European Union regulates banking-as-a-service through the most comprehensive framework of the four, and through a layered one. The European open-banking regime, established under the second Payment Services Directive and evolving toward its successor, governs the access that lets banking capabilities be consumed as services. Over it sits the Digital Operational Resilience Act, which, as Chapter 14 described, regulates the technology third parties on which financial institutions depend, banking-as-a-service platforms among them.12 Singapore regulates through the Monetary Authority of Singapore's technology-risk-management expectations and its framework for third-party arrangements, holding the licensed institution responsible for the partners it relies on.

The India Layer

India regulates banking-as-a-service through the accountability of the sponsor bank, and it has been tightening that accountability deliberately. The Reserve Bank's outsourcing guidelines establish the foundational principle that a bank remains responsible for the conduct of the third parties it engages. Through 2024 the Reserve Bank examined bank-fintech partnership arrangements closely and acted against specific arrangements it judged poorly governed, signaling that the sponsor role carries real and enforceable obligations.13 The 2025 Digital Lending Directions, examined in Chapter 12, govern the embedded-lending arrangements that run on banking-as-a-service rails. India's regulator, having watched the Synapse collapse from outside, has chosen to discipline the model before a domestic failure rather than after one.

The convergence across the four jurisdictions can be stated in one sentence. The United States enforces sponsor-bank accountability after a collapse, the European Union regulates the technology layer comprehensively and in advance, Singapore holds the licensed institution responsible through technology-risk expectations, and India tightens sponsor-bank accountability preemptively. Four regulators, four instruments, one shared conclusion: the licensed bank cannot delegate away its accountability for an embedded-finance stack, however many parties stand between it and the customer.

Practitioner's Lens: The Platform Deciding Its Embedded-Finance Strategy

Consider the head of strategy at a successful nonfinancial Indian platform, a quick-commerce application, a mobility service, a hospitality platform, or a software tool for small businesses, weighing whether and how to add embedded finance. The platform has customers, context, and data, the three assets Section 17.2 identified, and the chapter equips the practitioner to convert them into a strategy rather than a wish list.

The first discipline is to start from the customer's moment of need rather than from the catalog of available products. The platform should ask where, in its existing customer journey, a financial product would genuinely help the customer, a payment that removes friction, a loan at a point of real need, an insurance policy where a risk becomes salient. An embedded product placed at a genuine moment of need deepens the relationship; an embedded product bolted on because it was available irritates the customer and damages it. The context the platform owns is its single greatest asset, and the strategy must be built outward from the context.

The second discipline is to choose the banking-as-a-service partner with the Four-Party Stack framework in hand. The practitioner is selecting the party that will sit between the platform and a sponsor bank, and the selection should turn on the questions the framework raised: how is the ledger kept, can it be reconciled independently of the partner, how strong are the partner's controls, and how stable is the partner as a firm. The Synapse collapse fell hardest on the embedding platforms whose customers were frozen out, and an embedding platform that chooses its banking-as-a-service partner casually is choosing, casually, the firm that will hold its customers' money.

The third discipline is to be honest about accountability and brand. The customer will experience the embedded product as the platform's product, under the platform's brand, and will hold the platform responsible when something goes wrong, even though the regulatory accountability sits with a sponsor bank the customer never sees. The practitioner must therefore ensure that the platform's chosen partners are ones it would be content to be publicly associated with in a failure, because in a failure the platform's brand, not the bank's, is what the customer will name. Embedded finance lends the platform's brand to a financial product, and a brand lent is a brand at risk.

Applied Exercise

In this exercise the reader designs an embedded-finance strategy for an actual Indian platform. Select any nonfinancial Indian platform with a substantial customer base, a quick-commerce application, a mobility service, a hospitality platform, or a software tool used by small businesses.

Design the embedded-finance suite the platform should add over the next 18 months. For each product considered, payments, lending, insurance, and a card, decide whether the platform should add it, and justify the decision from the customer's moment of need rather than from the product's availability. For each product the platform does add, identify the banking-as-a-service partner and the sponsor bank, and apply the Four-Party Stack framework to show where the customer relationship, the ledger, and the accountability will each sit.

Set out the unit economics of the proposed suite, the revenue the platform expects to earn and the costs and risks it takes on, and the regulatory architecture, the licenses, the accountabilities, and the obligations the platform and its partners must satisfy. Present the work as an architectural note of roughly 3,000 words, and conclude with the single embedded-finance product you would launch first, and why.

Summary and Bridge

This chapter examined the architecture beneath digital banking: banking-as-a-service, the supply of banking capabilities as software, and embedded finance, the consumption of those capabilities by the nonfinancial platforms that now distribute financial products. It distinguished the supply side from the demand side, and it argued that the shift the two describe, finance moving from a destination to a feature, may matter more in the long run than the neobank story of Chapter 16.

The chapter offered the Four-Party Embedded Finance Stack as its framework, showing that an embedded-finance arrangement separates the customer relationship, the technical ledger, and the regulatory accountability across the embedding platform, the banking-as-a-service platform, and the sponsor bank, and that the separation is the model's power and its central danger. It used the Synapse collapse to show what happens when the middle party fails and the chain cannot be reconciled. It examined the Indian landscape led by M2P, Setu, and Decentro, the sponsor-bank question that determines which banks participate, and the demanding unit economics that make banking-as-a-service a hard business in which only the scaled and the disciplined survive.

The honest summary is that banking-as-a-service is genuine and necessary infrastructure for an embedded-finance era that is already arriving, and that it is a business unforgiving of weak controls, thin in margin, and dependent on a discipline of accountability that the Synapse collapse showed cannot be assumed. The firms best placed to endure, M2P among them, are those that reach scale, diversify their customers, and treat the reconciliation of the four-party stack as the core of the business rather than a detail of it.

This chapter and Chapter 16 examined the structures of digital banking, the neobank and the banking-as-a-service architecture beneath it. Chapter 18 closes Part 4 by turning from structure to the customer: the BankTech customer lifecycle, from acquisition through onboarding, engagement, cross-sell, and retention, and the question of what AI does, for good and for ill, at every stage of it.

Key Terms

Banking-as-a-service (BaaS). The packaging of banking capabilities into software that other firms can consume; the supply side of the embedded-finance market.

Embedded finance. The building of financial products into the customer experience of nonfinancial platforms; the demand side of the market.

Sponsor bank. In a banking-as-a-service arrangement, the licensed institution that supplies the regulated capability and carries the regulatory accountability for the whole stack; distinct from the UPI sponsor bank of Chapter 4.

Banking-as-a-service platform. The technology firm that packages a bank's capabilities into consumable software and operates the connecting layer, often including the ledger.

Embedding platform. The nonfinancial company that presents a financial product to its own customers, under its own brand, inside its own experience.

Four-Party Embedded Finance Stack. This chapter's framework: the chain of sponsor bank, banking-as-a-service platform, embedding platform, and end customer, across which the customer relationship, the ledger, and the accountability are separated.

Reconciliation gap. The failure, exposed by the Synapse collapse, in which the records held by a banking-as-a-service platform cannot be matched to the funds held at the partner banks.

Take rate. The thin per-transaction or per-account fee a banking-as-a-service platform charges; the basis of a volume-dependent business model.

Issuing-and-acquiring-as-a-service. The supply, as software, of the capability to issue cards and accounts and to process payments; the core of M2P's business.

Custodial deposit. An end customer's deposit held at a sponsor bank within a banking-as-a-service arrangement, the subject of post-Synapse recordkeeping reform.

Build-versus-buy decision. The choice an embedding platform makes between integrating a banking-as-a-service provider and building financial capability itself; the embedded-finance case collapses the Build-Buy-Partner choice of Chapter 7 to a two-way decision.

Sponsor-bank accountability. The regulatory principle that the licensed bank remains answerable for the conduct of every party in the embedded-finance stack it enables.

Discussion Questions

1. Is banking-as-a-service a sustainable business model, or a stepping-stone to either bank acquisition or platform collapse? Use Synapse, Solaris, and M2P to defend a position.

2. Should the Reserve Bank of India introduce a separate licensing category for banking-as-a-service platforms, distinct from the current treatment as outsourcing? Frame the arguments for and against.

3. Why has India's banking-as-a-service market produced infrastructure leaders such as M2P and Setu but no equivalent of Nubank, a customer-facing leader built on embedded finance?

4. Will agentic AI consolidate the banking-as-a-service layer around a few orchestrators, or fragment it further? Defend a view using the Four-Party Stack.

5. Apply the Four-Party Stack's diagnostic question, whether the chain can be reconciled if the banking-as-a-service platform vanishes, to an embedded-finance product you use. Can it?

6. The chapter argues that the mid-sized bank has more reason to become a sponsor bank than the largest banks do. If that is correct, what does it imply for the concentration of risk in the banking system?

7. Embedded finance lends the embedding platform's brand to a financial product. When that product fails, should the embedding platform or the sponsor bank bear the reputational and the legal consequence?

Further Reading

For the banking-as-a-service model and its risks, the press coverage and the bankruptcy-proceeding records of the Synapse collapse are the essential reading, because they document, in detail no summary can convey, exactly how a reconciliation gap forms and what it does to ordinary customers when it does.

For the Indian landscape, the public material of M2P Fintech, including its descriptions of its banking-as-a-service architecture and its international expansion, is the most useful source, and reading it with the Four-Party Stack framework in hand shows the banking-as-a-service platform's role from the inside.

For the regulatory dimension, the United States banking regulators' post-Synapse guidance and enforcement actions, the European Union's Digital Operational Resilience Act, and the Reserve Bank of India's outsourcing guidelines together show three regulators converging on the same principle of sponsor-bank accountability from three different starting points.

Finally, this chapter and Chapter 14 form a pair, one examining the infrastructure beneath lending and one the infrastructure beneath banking, and a reader who found the strategic conclusion of either compelling should read them together, because the embedded-finance era will be built on both layers at once.

Endnotes

  1. Reporting on the 2024 failure of Synapse Financial Technologies and the reconciliation gap between fintech-application records and partner-bank balances, which left end customers unable to access funds.

  2. M2P Fintech, disclosed scale: infrastructure powering more than 200 banks, more than 300 lenders, and several hundred further fintech engagements, serving end users numbering in the tens of millions.

  3. Reserve Bank of India scrutiny of bank-fintech partnership arrangements during 2024, including the March 2024 directions to Federal Bank and South Indian Bank to stop issuing new co-branded credit cards and the January 2024 order against Paytm Payments Bank (deposit bar effective March 15, 2024).

  4. M2P Fintech, financial disclosures: revenue of approximately ₹382 crore for the year ended March 2024, declining from ₹441 crore in the prior year, and more than ₹500 crore for the year ended March 2025, with the firm loss-making in both years, and company statements placing international markets at roughly 40 percent of revenue.

  5. M2P Fintech, Series D funding round of approximately ₹850 crore (on the order of 100 million dollars), announced September 2024, led by Helios Investment Partners, at a valuation surpassing ₹6,550 crore; and total funding raised on the order of 200 million dollars.

  6. Setu, acquisition by Pine Labs, 2022.

  7. Reporting on RBL Bank's fintech-partnership and co-branded-card business and the regulatory attention drawn to bank-fintech partnership arrangements.

  8. Reporting and bankruptcy-proceeding records concerning the 2024 collapse of Synapse Financial Technologies, including the shortfall between fintech ledgers and partner-bank balances, estimated by the Chapter 11 trustee at between $65 million and $96 million, and the prolonged loss of customer access to funds.

  9. Solaris SE, reporting on regulatory pressure, operational difficulties, and capital raises associated with its strategic reset, culminating in the €140 million financing round of early 2025 through which SBI Group became majority owner.

  10. Shopify Capital and Amazon Lending, public descriptions of merchant and seller financing underwritten on platform sales data.

  11. United States banking regulators, intensified supervision of bank-fintech partnerships following the Synapse collapse, including enforcement actions against partner banks and the FDIC's September 2024 proposed rule on recordkeeping for custodial deposit accounts with transactional features.

  12. European Union, the Payment Services Directive framework governing open-banking access, and the Digital Operational Resilience Act (DORA) governing technology third-party providers to financial institutions.

  13. Reserve Bank of India, Guidelines on Managing Risks and Code of Conduct in Outsourcing of Financial Services by Banks (2006), and Master Direction on Outsourcing of Information Technology Services (2023); and 2024 supervisory scrutiny of bank-fintech partnership arrangements.