Moving Money Was Never the Hard Part

As technology swept through one industry after another, the way we live changed beyond recognition. Long gone are the days of driving to Blockbuster to rent a movie; now we stream almost any film ever made, on demand, from anywhere on the planet.

Fintech, too, has spent years trying to fix the financial services industry through software and innovation. And yet one of the largest verticals in the world, arguably the largest, is still riddled with inefficiencies.

How is it that streaming a movie across the planet is instant, but sending money across a border still isn't much better than it was two decades ago? How are we still waiting three days for a transaction to settle in 2026?

SWIFT always takes the blame. It's every pro-stablecoin LinkedIn post’s punching bag, the culprit behind everything wrong with slow international payments. The stale incumbent that never innovated.

But SWIFT is just a messaging layer between banks. Blaming SWIFT for a slow payment is like blaming email for a contract that took three weeks to sign. The message was never the bottleneck.

The real bottleneck is the banks themselves. A cross-border payment rarely travels directly; it moves through correspondent banking - a chain of banks across different jurisdictions, each holding an account for the next, passing the instruction down the line until it reaches its destination.

And banks are heavily regulated for good reason. Every payment has to be checked to confirm it's legal. So the real time sink isn't the money in motion. It's what each bank does the moment a payment lands on its desk. Screen both parties against sanctions lists. Run the anti-money-laundering checks, and when something flags, open a case and put a human on it. Confirm the message carries all the required information, and if it doesn't, email the previous bank and wait for a reply. Debit one account, credit another. Then reconcile - make sure its own record of what it's holding matches the next bank's exactly, because the second two ledgers disagree, the payment stops until someone figures out why.

Now multiply that by every bank in the chain, each running different software, reading rules a little differently, staffing its own back office in its own time zone. The payment isn't sitting in a truck somewhere between countries. It's sitting in a queue, waiting for someone to clear an exception. The money could move in seconds. It's the work wrapped around the money that takes days.

That's the thing almost everyone misses about payments: moving money was never the hard part. The hard part is everything around the movement: compliance, reconciliation, record-keeping, the endless job of getting independent systems to agree on what just happened. Financial institutions spend over $206 billion a year on financial crime compliance alone. Behind every payment sits people downloading files, matching rows, and chasing exceptions over email.

We often confuse payments with settlement. But a payment is much more than the moment money moves.

Before the payment, there's intent. A contract, an invoice, a payroll run, a commission schedule. Something that says this money should move, for this reason, under these terms. Then comes execution, the movement itself: the payment routing across banks, processors, rails, and increasingly chains, each one carrying it a step closer while sanctions screening, monitoring, and approvals decide whether it's allowed to keep going. And after the payment, there's verification: reconciliation, ledger entries, fee checks, reporting.

A payment is only truly done when all three line up. When everyone involved agrees on what happened, and can prove it.

The catch is that no single system holds that full story. The intent lives in a CRM or an ERP. The execution happens across banks, processors, and rails. The verification sits in reconciliation tools, spreadsheets, and compliance platforms. Each of these was bought at a different time, from a different vendor, to serve a different team. And each one speaks its own language. Even two companies running the very same ERP will have configured it into different dialects.

So who stitches it together? People. Operations analysts pulling reports from five different portals. Finance teams doing semi-manual reconciliation because the automated one only catches part of the picture. Compliance officers piecing together context from systems that don't know each other exist. In most financial institutions, humans are the integration layer.

That mostly works. Until it doesn't.

Think about the last time an airline lost your bag. It rarely disappears mid-flight. It disappears at the transfer - the handoff between one airline's system and another's, where your bag briefly belongs to nobody. Each carrier's tracking works fine internally. The gap between them is where things break.

Payments fail the same way. Each individual system usually does its job. The failures happen in the handoffs, where no system is responsible, and no one is watching.

Synapse’s collapse is a clear example of this. Synapse sat between fintech apps and a network of partner banks, keeping the ledger of whose money was where. When it went bankrupt in 2024, the banks discovered their records didn't match Synapse's, and a court-appointed trustee identified a shortfall of $65 million to $95 million in customer funds. Not stolen. Just... unaccounted for, because four banks and one middleware company each held a partial view and nobody held the whole one. Thousands of ordinary depositors were locked out of their savings for months while forensic accountants tried to reconstruct reality from mismatched databases.

And this isn't a ‘startup’ problem. In 2024, Citigroup credited a customer account with $81 trillion instead of $280. Two employees reviewed the transaction and missed it; a third caught it 90 minutes after it posted. Citi reported 10 separate “near misses" of $1 billion or more that year. This is one of the largest, most heavily regulated, most technologically invested banks on the planet - and its controls still depend on a human noticing.

The promise of stablecoins is that they collapse all of this into a single shared ledger. No correspondents, no hops, no two banks reconciling their private versions of reality. Settlement that took days now takes seconds.

On the settlement leg, that's true - I've written before about why a universal ledger is a genuinely better foundation for moving money. Once every party is reading from the same ledger, there's nothing left to reconcile between banks, and one of the big reasons payments are slow simply disappears.

But stablecoins don't operate in isolation. They're just another rail, and in most cases they still have to interoperate with the off-chain rails everyone already uses. So institutions don’t experience blockchain as a simplification. They experience it as one more system that has to agree with all the others.

Take the typical stablecoin sandwich transaction. Dollars come in through a bank, get converted to a stablecoin at the on-ramp, move across a chain, get converted back to local currency at the off-ramp, and land in a bank account on the other side. That's at least three settlement environments, and multiple systems end to end once you count compliance, FX, and treasury.

Then layer on the operational mismatches. Chains settle 24/7; back offices work 9 to 5. On-chain finality is irreversible; fiat rails have chargebacks and recalls. Each rail has its own metadata, its own fraud patterns, its own audit expectations - which means each new rail spawns its own compliance workflow. And the regulatory bar is rising, not falling.

The math here is brutal. Connections between systems don't grow linearly - they grow combinatorially. Six systems that all need to agree with each other means fifteen pairwise relationships to keep in sync. Add a stablecoin rail, and you're at 21.

The obvious response in 2026 is to automate it. And honestly, the back office is where AI should shine. It's one of the most manual, labor-heavy functions in finance. Better yet, it's deterministic - a reconciliation either matches or it doesn't; a transaction either complies with policy or it doesn't. Unlike open-ended creative work, there's a defined right answer to check the machine against.

And the automation is coming. Every system in the stack will launch its own AI. Your ledger will get an AI ledger assistant, your compliance tool an AI analyst, your ERP its own copilot. But each of these only sees its own box. They inherit the exact problem we've been describing: no visibility across the systems that actually matter. An AI agent pointed at a single silo just automates that silo, faster.

So companies will try to go a level up and build agents that act across systems. This is the right instinct. However, every system speaks its own language, so a cross-system agent being fed raw data from a dozen systems that don’t agree on what a “settled payment” even means won’t be able to coordinate anything. And when a regulator asks what exactly your agents did, under whose authority, and whether you can prove it, you need an answer. If your systems can't produce that account for a human analyst today, they certainly can't produce it for software making a thousand decisions an hour.

So institutions can't responsibly automate this. It's too risky. Not, at least, until they have a layer that sits across all these systems: one that gives every agent a single, shared view, translates each system into a common language, and records everything that happens.

The industry doesn’t need another point solution that creates more complexity. It needs a layer that sits above the existing systems - without replacing them, and without touching the money. A system that does four things:

It's like the conductor of an orchestra. They don't play a single note. Every musician sees and plays only their own sheet music, and each one plays their part perfectly. But a part played in isolation isn't worth much. The magic is the whole thing together, and the conductor's only job is to hold all those parts in agreement, in real time, so what comes out is one coherent piece instead of a pile of correct but separate fragments.

This category doesn't have a settled name yet, but it's forming. Cordant, built by an ex-Rapyd team, came out of stealth recently around exactly this thesis, backed by an $8 million seed and early partners like Bitso and Paxos.

The last few years of fintech have been all about moving money faster and cheaper. Digital assets are real now, and blockchain rails are on their way to becoming a commodity, available to everyone, no longer a differentiator.

The future belongs to whoever can govern the movement. The institutions that can see across their systems, catch the gaps in real time, and prove why every decision was made will onboard partners faster, adopt new rails without adding headcount, and deploy AI while their competitors are still writing memos about it.

The ones that can't will keep discovering their problems the way Synapse's depositors did: after the harm is already done.

(The views and opinions expressed in this article reflect the independent analysis and personal perspectives of the author. They do not represent the official positions or endorsements of their employers or this publication.)