By Heorhi Tratsiak
There is a moment that stays with me from my early years at the bank.
A woman in her sixties sat across from me at the client desk, a folded piece of paper in her hands. Her son was working in Poland, she explained. He had sent her money three weeks ago. The transfer had not arrived. She had called the bank twice. She had been told to wait. She was not angry, exactly. She was confused in the way people get confused when something that seems simple turns out to be invisible.
I processed the inquiry. I traced the transfer through our internal systems, found where it had stalled, flagged it for the international payments team. The money arrived in her account four days later. She came back to thank me, which clients rarely did.
What I did not tell her, because there was no version of it that would have helped, was how many institutions her son’s money had actually passed through. How many systems had touched it, held it briefly, checked it, passed it forward. How many countries it had technically visited before arriving in Minsk. She thought her son had sent her money. In the way most people understand the word “sent,” he had not. What he had sent was a message. What moved was a promise.
I spent seven years inside a financial institution that was part of one of the world’s largest energy conglomerates. I worked with SWIFT, with Visa and Mastercard, with Apple Pay and Samsung Pay, with correspondent banks across Europe and beyond. And the thing that struck me most, consistently, across all those years, was not how sophisticated the system was. It was how invisible.
The global financial infrastructure processes trillions of dollars every single day. It touches every person on earth who has ever received a salary, paid a bill, or bought anything with a card. And almost no one outside the institutions that operate it has any clear picture of what it actually is, how it works, or where its real pressure points are.
That invisibility is not an accident. Understanding the architecture is the first step to understanding everything else about money in the modern world: why sanctions work, why they sometimes don’t, why fintech has struggled to truly disrupt banking despite two decades of trying, and what it will actually take to build something different.
What Money Is Not
When you swipe your card at a coffee shop, you do not send money anywhere. Nothing moves. No funds travel from your account to the merchant’s account at the moment of transaction. What happens instead is that several institutions exchange messages with each other at very high speed, and each of them updates a number in a database.
This is not a minor technical detail. It is the fundamental nature of how modern money works, and almost no financial education explains it this way.
Money, in the contemporary sense, is not a thing. It is a record of an obligation held by an institution. When you have a thousand dollars in your account, what you actually have is a claim against your bank. The bank owes you that amount. It does not have a thousand physical dollars sitting somewhere with your name on it. It has its own records, its own obligations to other institutions, its own assets that may or may not include cash in any physical form.
When your bank owes you money and you want to transfer some of that claim to someone else, a chain of institutions needs to update their records to reflect that the obligation has moved. Your bank reduces its obligation to you. The receiving bank increases its obligation to the recipient. For this to work, there needs to be a shared language between institutions, a system for verifying that messages are authentic, and a way to settle the net flows between banks at the end of each day.
The global financial system built over the past century is, at its core, a solution to those three problems.
SWIFT: The Postal System That Runs Global Finance
Most people who have wired money internationally have encountered the term SWIFT, usually as an acronym on a bank form alongside a long string of letters and numbers. Very few people understand what it actually is.
SWIFT, which stands for the Society for Worldwide Interbank Financial Telecommunication, is not a payment system. This distinction matters enormously. SWIFT does not hold money. It does not process transactions. It does not settle funds. What SWIFT does is carry messages between financial institutions in a standardized, secure format.
When your bank in one country wants to send instructions to a bank in another country, it formats those instructions according to SWIFT standards and sends them through the SWIFT network. The receiving bank reads the message, understands what it says because SWIFT messages follow strict protocols, and acts on it. The actual movement of funds happens separately, through a parallel system of correspondent banking relationships that we will come to shortly.
Founded in 1973 and headquartered in Belgium, SWIFT operates as a cooperative owned by its member financial institutions. It connects over eleven thousand financial institutions in more than two hundred countries. At peak times, it carries more than forty million messages per day.
Inside the bank where I worked, a SWIFT terminal was as fundamental to operations as the electricity. International transfers, trade finance instructions, currency exchange confirmations, letters of credit: all of it flowed through SWIFT. The messages themselves followed specific format types. MT103 for a customer credit transfer. MT202 for a bank-to-bank transfer. MT700 for a letter of credit. The formats had been standardized over decades, which is part of why the system works across such a vast number of institutions in such varied regulatory environments.
What I understood from working inside this, that you cannot fully appreciate from outside, is that SWIFT is not just a technical network. It is a trust network. Institutions communicate through it because membership in SWIFT carries implied verification. To be connected to SWIFT means your institution has been vetted, that you operate under some regulatory framework, that messages arriving from your institution carry a baseline level of credibility. The technical system and the trust system are not separable.
This is precisely why disconnecting an institution or a country from SWIFT is such a significant event. It is not just cutting off a messaging service. It is removing that institution from the web of trusted counterparties that the global financial system runs on. It is, in a very practical sense, making that institution invisible to the rest of the financial world.
The Correspondent Banking Web
SWIFT carries the messages. But messages alone do not move value. For actual funds to flow between institutions in different countries, those institutions need accounts with each other, or with a chain of intermediaries who do.
This is the correspondent banking network, and it is older than SWIFT, older than electronic communication, older than most of the countries that participate in it. The basic principle dates to the Renaissance-era Italian banking houses that maintained agents in trading cities across Europe, allowing a merchant in Florence to draw funds in Bruges without physically transporting gold.
The modern version works like this. Bank A, located in Belarus, needs to send dollars to Bank B, located in Japan. Bank A and Bank B do not have a direct relationship. But Bank A has a correspondent account at a major American bank, let us call it Bank C. And Bank B also has a correspondent account at Bank C, or at Bank D which has an account at Bank C. The message travels through SWIFT. The actual dollar movement happens through Bank C updating its internal records: decreasing the balance it holds for Bank A, increasing the balance it holds for Bank B.
The dollars, in some literal sense, never left the United States. They moved between accounts at an American bank. What changed were the ownership records.
The implications of this are not abstract. Dollar-denominated transactions, regardless of where they originate and where they are destined, must at some point pass through an institution that holds accounts at the Federal Reserve. This is because the US dollar settlement happens on Fedwire, the Federal Reserve’s settlement system. An institution that loses its correspondent banking relationships, that finds no American bank willing to hold its accounts, effectively loses the ability to process dollar transactions. Not just with American parties. With anyone, anywhere in the world, who wants to settle in dollars.
Euro transactions follow the same logic but through TARGET2, the Eurosystem’s large-value payment system. Sterling through CHAPS. Each major currency has its own settlement infrastructure, and access to that infrastructure requires relationships with the institutions that are connected to it.
During my years in international payments, I spent a significant amount of time working within these correspondent relationships. Processing international transfers meant understanding not just the SWIFT messaging format, but the specific requirements of our correspondent banks. What documentation they needed. How they would handle transactions from certain jurisdictions. Where they would apply additional scrutiny. The relationship with each correspondent bank was its own landscape, with its own rules, its own friction, and its own particular sensitivities.
What It Looks Like From the Inside
There is a version of the correspondent banking world that exists in academic papers and central bank publications, and then there is the version that exists at the desk of someone processing international transfers on a Tuesday afternoon.
The academic version describes correspondent banking as an efficient mechanism for global value transfer. The operational version is messier, more human, and more instructive.
Each correspondent relationship carried its own history. The large Western bank we used for dollar clearing had onboarded us years before I joined, when the regulatory environment was different and the documentation requirements were lighter. Maintaining that relationship required periodic reviews, updated compliance questionnaires, occasional requests from their compliance team to explain transaction patterns that seemed unusual from their perspective.
I learned early that compliance questions from a correspondent bank were not abstract bureaucratic exercises. They were signals. A pattern of questions about a particular category of transactions meant their compliance team was watching that category. A sudden increase in required documentation for transactions from a specific jurisdiction meant something had changed in the risk environment as their regulators saw it. Reading correspondent bank behavior was a form of intelligence gathering about the regulatory priorities of the financial systems on the other side.
Processing a payment that appeared problematic in some respect required judgment that no rulebook fully covered. The amount was large but not unusual for this client. The counterparty was in a jurisdiction that was technically acceptable but that our correspondent had recently asked questions about. The documentation was complete, but something about the transaction pattern felt off. These were the decisions that occupied most of my working attention, not the technically complex transfers, which largely processed themselves once set up correctly, but the boundary cases.
What I observed over seven years was that the institutional rules governing these decisions were always incomplete. They set the outer edges but left significant space in the middle. And in that space, a great deal depended on the judgment, attention, and implicit knowledge of the people making the decisions. The system looked systematic from the outside. From inside, it was sustained by accumulated human expertise in ways that the formal documentation never fully captured.
When senior colleagues retired or moved to other positions, specific knowledge walked out with them. The new person might follow the same procedures and produce different outcomes, because the procedure did not carry the understanding of when to apply it strictly and when to use discretion. Banks talked constantly about systemizing this knowledge, building it into processes, making it transferable. In seven years I watched this aspiration collide repeatedly with the reality that the most valuable knowledge was the kind that was hardest to write down.
This is worth keeping in mind when evaluating claims that artificial intelligence will systematize financial compliance and remove the human judgment problem. The problem is not that humans are inconsistent. It is that the situations requiring judgment are inherently ambiguous, that the rules governing them are always partially incomplete, and that good judgment requires not just applying a framework but knowing when the framework is insufficient for the situation in front of you. Making that call is not a technical problem. It is a problem of experience, institutional knowledge, and the particular kind of intelligence that develops through years of doing something in a specific context.
The infrastructure handles the routine. The humans handle the edge. And in financial systems, the edge is where almost everything that matters actually happens.
The Card Networks: A Different Kind of Infrastructure
Parallel to the SWIFT and correspondent banking world, there exists a separate but equally invisible infrastructure that handles the billions of everyday consumer transactions that happen when people pay by card.
Visa and Mastercard are often described as payment companies. This is understandable shorthand, but it misrepresents what they actually are. Neither Visa nor Mastercard moves money. Neither holds your funds, processes your transactions in the sense of transferring value, or takes on credit risk when you make a purchase. What they operate are networks of rules and messaging standards, together with the clearing and settlement infrastructure that connects banks around the world.
The actual model involves four parties, and understanding it changes how you see the industry.
When you use your Visa card at a shop, there is the issuing bank (the bank that gave you the card), the acquiring bank (the bank that serves the merchant), the card network (Visa or Mastercard), and the merchant. At the moment of transaction, the acquiring bank’s terminal sends a message asking whether your issuing bank authorizes the charge. That message travels through Visa’s network. Your issuing bank checks your available funds, applies its fraud models, decides yes or no, sends back an authorization message through the same network. The whole exchange takes under two seconds. The merchant gets confirmation. You get your coffee.
What does not happen at that moment is any movement of actual funds. Authorization and settlement are separate processes. The actual transfer of money from your bank to the merchant’s bank happens later, typically overnight, through a batch process that nets out all the day’s transactions. Visa sits in the middle, as the shared infrastructure that makes it possible for thousands of issuing banks and millions of acquiring banks to interoperate without needing bilateral agreements with each other.
My specific work at the bank placed me at the interface between these network rules and their operational implementation. When Visa or Mastercard updated their technical specifications, I worked on making sure our systems reflected those changes. When new payment technologies were introduced, things like contactless payments, Apple Pay, Samsung Pay, I was part of the team that worked through the certification requirements.
What certification for these services actually involves is less well understood than it should be. The public narrative tends to present bank-tech partnership as a straightforward commercial relationship: tech company offers product, bank signs up, customers get access. The operational reality is considerably more involved. Each new payment method carries its own tokenization requirements, its own security protocols, its own liability rules for disputed transactions. The bank’s systems need to be updated to handle the new data formats. The bank’s staff need to understand the new dispute resolution process. The compliance documentation needs to be completed. And all of this needs to be verified by the network before the bank is certified to offer the service.
This certification process has consequences that matter beyond any individual institution. It creates a baseline of quality control across the network. It also creates significant friction for anyone trying to enter the ecosystem. A fintech company that wants to issue Visa cards needs a licensed bank to sponsor their program. That bank needs to certify not just its own systems but also the fintech’s systems, under its own regulatory umbrella. The bank carries responsibility for the fintech’s compliance. This is why many fintech partnerships with banks take twelve to eighteen months before a product launches. It is not bureaucratic laziness. It is the network enforcing a standard that protects its own integrity.
Where the Power Actually Sits
I want to say something directly about what working inside this system taught me about power, because it is not immediately obvious from the outside.
The financial infrastructure is not neutral. It never was, though it was designed and presented as if it were. A neutral technical system does not need to be owned by anyone. A neutral technical system does not have members and non-members, does not have standards controlled by specific institutions, does not have settlement processes that concentrate through the central banks of a handful of countries.
SWIFT is incorporated in Belgium and governed by a board that represents its member institutions, which are heavily weighted toward Western financial institutions. The dollar settlement system runs through the Federal Reserve. Visa is an American company. Mastercard is an American company. These are not coincidences of history. They are the accumulated result of decades of financial and geopolitical decisions that embedded these institutions at the center of global commerce in a way that created enormous leverage.
That leverage is most visible when it is exercised.
When a country or institution is cut off from SWIFT, the immediate practical effect is the loss of the ability to send or receive standardized international payment messages. But the downstream effects go much further. Without SWIFT access, correspondent banking relationships deteriorate because banks in other countries have no reliable way to communicate with the sanctioned institution. Without correspondent banking relationships, dollar settlement becomes impossible. Without dollar settlement, international trade becomes extremely difficult, since the dollar remains the dominant currency for global commodity pricing and trade finance.
The Iranian sanctions that began with SWIFT restrictions in 2012 reduced Iran’s oil exports by roughly forty percent within a year. Not because anyone physically prevented oil from being pumped or shipped. Because the financial plumbing required to get paid for that oil was no longer accessible.
Inside our bank, I watched a smaller version of this logic play out in real time. When Western sanctions expanded following the geopolitical events of early 2022, the effects arrived not as a single dramatic cutoff but as a gradual tightening. Correspondent banks that had previously processed our transactions without comment began applying additional scrutiny. Some transactions that had moved smoothly for years started taking longer, or bouncing back with requests for documentation that had not previously been required. Certain services became unavailable. The network, as a network, was withdrawing.
This is how financial infrastructure as a tool of pressure actually works at the operational level. It is not usually a single switch being thrown. It is a hundred small decisions by a hundred institutions, each acting in their own compliance interest, that collectively produce an environment in which the sanctioned entity finds the normal business of banking becoming progressively more difficult.
Understanding this mechanism from the inside is quite different from reading about it in a geopolitical analysis. The geopolitical analysis tells you that financial infrastructure is being used as leverage. Working inside a bank under that pressure teaches you that the leverage is not centralized. It is distributed. It is alive in every correspondent bank relationship, every card network certification, every correspondent account balance. The infrastructure does not turn against you all at once. It turns against you incrementally, through a thousand small friction points, each individually defensible as compliance procedure and collectively devastating.
The System’s Dependence on Nodes
Networks are not equally distributed. Some nodes carry more traffic than others. Some nodes, if removed, would degrade the network significantly. Others could disappear without much practical consequence.
The global financial messaging network is no different. The concentration of dollar-clearing through a small number of major American institutions, the concentration of euro-clearing through the European Central Bank infrastructure, the concentration of SWIFT governance in Western institutions: these create specific points where the network can be influenced by those who control or have access to those points.
De Nederlandsche Bank estimated that roughly sixty percent of all international dollar transactions pass through just three American banks. This is not a vulnerability that anyone designed into the system. It is the natural result of correspondent banking economics: volume begets lower fees, lower fees beget more volume, and the market consolidates around the largest players. But it means that the effective reach of American financial policy extends well beyond the institutions that American regulators formally supervise.
Any institution that needs to process dollar transactions, which is to say almost every financial institution in the world, is in a relationship of dependence with the institutions that sit at those nodes. Maintaining access to dollar clearing requires maintaining relationships that those node institutions are willing to maintain. Which means operating in ways they will find acceptable. Which means being sensitive to the preferences of American financial regulators, because those regulators are the institutions’ primary risk.
This is what is usually meant, imprecisely, when people talk about the “weaponization” of the dollar. The dollar itself is not a weapon. What has been used as a tool of policy is the infrastructure through which dollar transactions flow, and the leverage that comes from controlling or having access to the nodes through which the global system depends.
The sophistication of this mechanism is worth sitting with for a moment. It does not require passing laws in every country. It does not require bilateral agreements with every government. It operates through the normal compliance behavior of private institutions, each acting rationally in their own interest, each simply not wanting to take on regulatory risk by maintaining relationships that their own supervisors would scrutinize. The geopolitical effect is produced by market behavior. The hand that moves it does not need to be visible.
What the System Cannot See
The architecture I have been describing is extraordinarily powerful, but it is not complete. It has limits. And those limits are as important as the capabilities.
The system was designed for a world in which the most significant financial actors were large institutions. Banks. Trading houses. Governments. Corporations with formal legal existence and regulatory relationships in recognized jurisdictions. The SWIFT model, the correspondent banking model, the card network model: all of these assume counterparties with identities that can be verified, accounts that can be monitored, legal persons against whom sanctions can be enforced.
Cryptocurrency, in its original conception, was a direct challenge to these assumptions. A transaction between two individuals who hold cryptographic keys does not require a bank. Does not require a SWIFT message. Does not pass through a correspondent. Does not settle at the Federal Reserve. It settles on a distributed ledger that no single institution controls.
The financial establishment’s response to cryptocurrency has been, broadly, to assimilate it at its edges while containing it at its core. Exchanges that convert cryptocurrency to fiat currency are regulated. They apply know-your-customer requirements. They report suspicious transactions. They can be sanctioned. The anonymous layer of cryptocurrency can be used to move value, but to get that value into the conventional financial system, where it can buy real goods and pay conventional taxes, it still has to pass through regulated exchanges that are subject to the same pressure points as any other financial institution.
This has not eliminated cryptocurrency’s utility for sanction evasion. It has reduced it and made it more complicated and more expensive. The fundamental challenge the technology poses to the existing architecture remains unresolved.
China’s development of the digital yuan represents a different kind of challenge: not a decentralized challenge, but a state-backed one. A payment infrastructure that settles in a different currency, through different institutions, with different regulatory oversight, potentially extending its own network of bilateral relationships with countries that would prefer not to be dependent on dollar-clearing infrastructure. This is not yet a serious alternative to the existing system for most of global trade. It is an attempt to build the infrastructure of an alternative, country by country, relationship by relationship.
These are the developments worth watching if you want to understand where the architecture of global finance is heading. Not the technology in isolation, but the geopolitical incentives that will shape how it develops and who it serves.
Why It Matters to Understand This
I began this piece with a woman waiting for her son’s money to arrive. She was not thinking about correspondent banking relationships or SWIFT messaging formats or the Federal Reserve’s settlement infrastructure. She was thinking about her son, and about paying her bills.
Most of the people who interact with the financial system every day never need to think about its architecture. It is, in the best case, invisible infrastructure: present, reliable, not requiring attention. The woman I helped was experiencing the rare case where something went wrong, and the invisible became briefly visible as an obstacle.
But the invisibility of this infrastructure has a cost beyond individual inconvenience. When the infrastructure is used as an instrument of economic policy, most of the people affected have no framework for understanding what is happening to them, or why. When financial technology companies promise to transform banking, most of the people receiving those promises have no basis for evaluating whether the transformation is as fundamental as it appears, or whether it is a new interface sitting on top of a structure that remains unchanged.
And when the architecture itself begins to shift, as it is beginning to shift now, with alternative payment networks under development, with central banks experimenting with digital currencies, with the correspondent banking network consolidating further, the people who will shape those changes need to understand what the current architecture actually is, not the simplified version that gets described in press releases and introductory economics courses.
I spent seven years working inside this architecture. What I observed is that its technical complexity is real but manageable, that its power comes as much from network effects and institutional trust as from any technical feature, and that it is far more politically embedded than its presentation as neutral infrastructure would suggest.
The next question, which will take the rest of this series to work through, is what happens when that infrastructure is stress-tested. When politics enters the system not as a distant background condition but as an immediate operational reality. When the normal flows of the network are disrupted, and the people working inside its institutions have to figure out in real time what that means.
The invisible architecture becomes visible under pressure. What you see, when it does, tells you everything about how the system actually works.
That is what the next chapter is about.
Heorhi Tratsiak spent seven years at Belgazprombank, working with international payment systems including Visa, Mastercard, Apple Pay, Samsung Pay, and SWIFT. He is the founder of TSG Tech Ltd, incorporated in England and Wales, and writes about financial infrastructure, AI, and the economics of technology. Follow him on Medium and LinkedIn.
