Programmable Finance: What the Convergence of AI and Stablecoins Means
And which questions you should be asking yourself right now
Having spent more than two decades in Payments, I know one thing for sure…It never gets boring. Every few years, something happens that seems to shift the entire industry in a different direction.
Not so long ago, we debated how the commoditization of payments is putting pressure on the take rates of PSPs. Still, in just the span of a few months, we are no longer talking about payments being the plumbing but rather about how it is turning into the realm of a genuine platform shift.
This week’s Stripe Sessions felt like one of those moments.
Stripe showed an AI model built on “tens of billions” of transactions that pushed its card-testing catch rate up by 64 per cent overnight (Stripe). Minutes later, the same keynote team announced Stablecoin Financial Accounts, regulated, always-on dollar balances available in 101 countries (Ledger Insights).
A model that thinks about risk and a token that settles against it in real time: that pairing gives payments a new centre of gravity, and it explains why Visa, Coinbase, PayPal, Circle, and even the European Central Bank all surfaced related projects inside the same quarter.
In this edition of the Payments Strategy Breakdown, I will share why money is beginning to behave like software, and the companies that already move fast in code will decide how that software gets used, and the questions you should be asking to decide what to do next.
Let me explain…
What “programmable” really means
For thirty years, we optimised form factors; swipe to chip, chip to contactless, and plastic to phone.
Programmable finance shifts the work to two new layers:
Inference at the edge. Foundation models turn raw authorisation data into live probability scores. Fraud, credit, and compliance decisions no longer wait for rule-based back-office engines.
Tokenised settlement. Fully-reserved stablecoins or central-bank pilots close the gap between authorisation and clearing. Cash flow is not “T + 2”; it is block-height + 1.
Run together, those layers mean a payment rail can examine intent, price risk, and move value without opening a browser window or firing a batch file.
That, in a nutshell, is what Stripe, Visa, and Coinbase are now testing in public.
What have they been doing?
Stripe sets the pace.
Beyond the fraud lift, Stripe’s AI model now tunes authorisation logic for each issuer; early pilots show several-point increases in approval on cross-border traffic.
Its new stablecoin accounts hold USDC or USDB alongside fiat balances, and the firm has already routed small-value supplier payouts over Polygon to prove the latency win (Ledger Insights).
NVIDIA’s migration of every GeForce Now subscription within six weeks underlines just how quickly a motivated enterprise can swap engines.
Visa rewires the network.
Through Bridge (acquired by Stripe in January), Visa now lets fintechs issue cards that debit a stablecoin wallet and spend at any of the 150 million scheme merchants (Visa).
Settlement still clears through VisaNet, but funding can happen on-chain, and developers reach it through one API.
A solution that quietly dissolves the historical line between on-and-off-network value stores.
Coinbase pushes payments into HTTP.
The new x402 proposal resurrects the long-dormant 402 Payment Required status code and attaches an instant-settlement USDC transfer to it (Coinbase).
In a single curl command, an AI agent or a SaaS invoice bot can request a resource and pay for it at the protocol layer.
If x402 gains traction, clearing could disappear from the application stack entirely.
What else…
At the same time, PayPal is leaning on its own PYUSD for enterprise flows; Circle launched its own cross-network settlement hub; the European Central Bank is prototyping an offline digital euro with conditional logic baked in; and the Canton consortium wants the same idea for tokenised securities.
In other words, different motives, but a similar architecture, which is focused on combining deterministic tokens with dynamic AI policy.
But where does that leave the rest of us?
Having been in several large Payments companies myself, and having worked with dozens of Payments companies trying to catch up with the leaders in the industry, I know that every single C-level team, is currently getting their best and brightest to draft a pitch deck with ideas to do exactly the same as Stripe, Coinbase, Visa and Circle.
But before we do that, let’s figure out what the conversation should be about, before we come up with new ideas.
Let’s understand the implications first.
In this instance, it will be the finance chiefs who will recognise the obvious upside of these announcements.
Less float, lower FX, and potentially higher approval.
The more challenging part is risk ownership.
A stablecoin balance makes the treasury responsible for smart-contract events and de-pegs. An AI score that declines a transaction moves from a deterministic rule to a statistical judgment; regulators will ask about bias and explainability.
The practical hurdles sit closer to home:
Legacy gateways that still roll up device data once per night cannot feed an inference endpoint that expects millisecond context.
On-chain liquidity management needs engineers who can audit solidity code and back-test stress scenarios, not just reconcile Nostro accounts.
Compliance teams must map PSD3, MiCA, and the forthcoming US stablecoin bill onto token flows that PSD2 never covered.
The win is clear enough to start the debate, but the implementation work crosses risk, engineering, and policy lines that many PSPs still keep separate.
So, how do you then manage risk and reward?
Think of the opportunity as a two-dimensional spread:
High reward / manageable risk: Stripe already owns the data and model talent; a controlled, closed-loop stablecoin account adds yield and speed.
High reward / high risk: Visa’s global footprint gives scale, but it must harmonise smart-contract execution across more than 200 jurisdictions.
Modest reward / low risk: A regional acquirer that exposes stablecoin pay-outs in one corridor boosts margin and stays inside existing licences.
Low reward / high risk: Any PSP that launches tokens before it can track exposure per address will learn that treasury events cascade faster than chargebacks.
Boards should decide where on that grid they stand before engineers carve out a sprint budget.
Seven questions every payment company should answer this quarter
But how do you start the conversation? What are the questions that you should be asking yourself and your team, on how to respond to these developments in the market?
Here are a few of the questions that I have been using to help C-level executives think about Programmable Finance and how the convergence of AI and Stablecoins could impact their business.
Data readiness: Do we have the tools and data to keep our fraud-detection system learning from every new transaction, or will we need an outside provider to handle that for us?
Rail arbitration: Do we even know which transactions would be cheaper or faster on a token rail today?
Liquidity comfort: Who takes the call if USDC trades at 0.98 for an hour on a Friday night?
Reg scope: Which regulators need advance notice if we move from e-money to fully-reserved stablecoins?
Customer promise: What is the single metric, speed, approval, or FX cost, that we will improve by at least 10 per cent?
Stack resilience: How many flows still rely on nightly batch files and what does that really cost us?
Talent gap: Do we employ even one engineer who can read both Solidity and the latest fraud-model audit report?
Answer those, and you will have a good starting point on knowing whether to build, partner, or wait.
Where is this heading
What I am noticing is that innovation cycles in payments continue to compress the more we move into a pure-software path.
Tokenisation took half a decade; Tap-to-Pay moved in three years.
AI-scored, token-settled payments are software from end to end, so the adoption curve will look steeper still.
Incumbency helps, until it doesn’t.
Global processors have brand and scale, but focus, agility, and regional expertise often convert faster when the ground shifts.
Over the next two years, look for mid-sized payment companies in Latin America, Africa, and Southeast Asia to outpace larger rivals by skipping the traditional card networks and letting AI choose the best path for each transaction in real time.
If your platform can already decide in real time where to send a transaction, adding a token rail is a configuration change. If it can’t, the more pressing project is modernising the decision layer, not debating blockchain.
Programmable finance will reward the teams that ship proofs of concept before summer ends and treat model risk, treasury, and developer experience as a single roadmap item. The field is still open, and speed, not size, will decide the early winners.
Thank You for Reading. Please like, Comment, Share, or Post on Your Social media. I appreciate all the feedback I can get.
P.S. If you want to work with me in a larger capacity, either speaking, advisory, or consulting, feel free to email or DM me.