Original · Unique Research · 2026-04-01 · Shanghai
Editor's note: This historical article presents the author's commentary and interview answers attributed to Sean Li, identified by the source as Airwallex's Billing product lead. License counts, country coverage, bank partnerships, capabilities and performance figures are source-reported company claims, not independently verified here or statements of current availability. The approximately 40% improvement has no supplied baseline or calculation and is not translated into a claim that elapsed time fell by 40%. T0 finance and the 12-month/end-of-2026 outlook remain plans and forecasts, not completed outcomes. This is not financial, legal or compliance advice; human approval and responsibility are retained. Sixteen source images still require content review.
Unique Awards
What AI Companies Going Global Really Lack
Is Not Just Payment Tools, but a Financial Operating System
Actually Collect the Money, Manage It and Move It Safely
When people discuss AI companies expanding internationally, models, growth and breakout products are the most visible things.
Who launched on Product Hunt, who built a subscription business in the United States, whose Agent took off in overseas communities—discussion often stops there.
But once you put a company into the reality of global operations, you discover that the hardest problems are often in the back office, not the front.
Users have arrived—how do you collect their money? If you can collect in North America, can you collect in Europe? Once subscriptions take off, how do you bill them? With multiple operating entities, how do you consolidate funds? How do you compliantly pay for compute, advertising, employees, contractors and service providers? What happens when risk controls wrongly block legitimate activity? When KYC, AML, data-residency or foreign-exchange rules change, who handles it?
These questions attract far less attention than “the model has improved again,” but they often determine whether a company can move from “having a product” to “having a business.”
That was also my strongest impression after interviewing Sean Li, Airwallex's Billing product lead: for an AI company going global, the real difficulty is not selling the product, but actually collecting the money, managing it and moving it safely.
Cross-Border Finance Used to Be a Patchwork of Tools. Now It Is Becoming a System.
Sean Li repeatedly emphasized an important point in the interview: what AI businesses most fear today is not a lack of tools, but excessive fragmentation.
A global company previously often had to assemble many providers itself: a PSP, banks, foreign-exchange services, corporate cards and expense-management systems, then connect them manually or through an internally built system. The result was fragmented data and disconnected processes, leaving even AI with nowhere to start.
That captures the reality for many AI companies today:
• The product is AI native,
• Growth is AI native,
• Marketing is AI native,
• But finance and treasury systems often still use the old methods.
At first, that is manageable. With one market, one currency and one entity, people can still carry much of the load. But once global subscriptions, cross-region acquiring, multi-entity operations and global payouts begin, the problem can rapidly go from “a little troublesome” to “a complete bottleneck.”
Airwallex therefore positions itself as more than a cross-border payments provider: it combines accounts, foreign exchange, acquiring, cards, financial software and AI into an AI-native financial OS. Sean Li said its business centers on the cross-border financial needs of globally operating companies. Drawing on what he described as 87 financial licenses and a payment network spanning more than 200 countries and regions, it serves cross-border e-commerce, foreign trade, gaming and travel, while making a particular push into AI companies' international expansion.
The logic behind this is simple:
as business operations become more global, more real-time and more automated,
financial capabilities can no longer remain isolated features;
they must become a continuous system.
What AI Companies Most Need Is Not “One More Payment Interface,” but Fewer Providers to Stitch Together
One detail in the interview illustrates the problem especially well.
Sean Li said Airwallex's global billing-management product Billing, built for AI companies expanding overseas, does not simply handle one point of payment collection. It integrates multi-currency settlement, automated reconciliation and intelligent billing into one system. Without such a system, he said, an AI company would need to connect at least 3 to 5 providers.
That is a persuasive observation. It explains why AI companies may need this infrastructure more than traditional businesses—not because they like new technology more, but because they have more to fear from fragmentation.
AI companies usually have several characteristics.
First, they are born global. Many products serve developers and business users worldwide from launch, rather than starting locally and gradually expanding.
Second, their charging models are flexible. Rather than a single monthly fee, subscriptions, trials, renewals and usage-based charging often coexist. Sean Li noted that Airwallex Billing addresses precisely these needs: AI-business invoices, recurring subscriptions and complex usage-based billing.
Third, their organizations are lean, but move quickly. Their teams may not be large, yet they soon encounter global payment collection, compute procurement, overseas hiring and contractor payouts. Sean Li also said Airwallex hopes to become standard financial infrastructure for AI companies, covering global subscription billing, compute procurement and payments to employees and contractors worldwide.
For these companies, the question is never simply “are there tools?” It is “can the tools connect to one another?”
AI Has Accelerated Payments, but Made Compliance More Demanding
One of my favorite lines in the interview was this:
AI has brought F1 speed to execution,
but made the design, compliance and governance layer
more demanding.
That puts the issue clearly. Discussions of AI's impact on finance today often focus only on “what has become faster.”
Take account opening and KYC review. Sean Li said generative AI and stronger contextual understanding help reduce false positives and shorten reviews without lowering risk-control standards, improving overall account-opening and onboarding turnaround by approximately 40%.
Or take intelligent risk controls and payment routing. In the source's account, AI can choose routes, make intelligent retries, decide whether to trigger 3DS and score fraud behind the scenes, determining in milliseconds which channel to use, whether 3DS is needed and whether a transaction is high-risk. For high-frequency subscription AI/SaaS businesses, this directly affects payment success and chargeback rates.
Then there are customer service and product guidance. The interview describes AI Assistant as doing more than answering questions: it can trigger actions within secure, controlled boundaries, such as opening accounts, issuing cards, enabling payment methods and checking incoming and outgoing payment status, bringing previously scattered processes into a single conversation window as far as possible.
Those are the things that have become “faster.”
But two other layers have become more demanding.
One is upfront risk-control and compliance design and rule-setting. Which scenarios can AI clear automatically, and which require “AI recommendation + human approval”? Who is responsible for a wrong decision, how is it rolled back, and how is an audit trail preserved? This layer is more complex than in a purely manual era.
The other is data and interface engineering. If account, transaction, expense, risk-event and customer data are not modeled consistently and exposed through an API, an Agent cannot see the whole picture, let alone make genuinely intelligent decisions.
That is a practical rule of the AI era: the further AI moves into execution, the less the underlying system can afford to be a patchwork.
In the past, some system fragmentation might merely reduce efficiency. Today, it can create risk exposure.
What Is Really Hard to Copy Is Not a Feature, but the Underlying Network Already Built
Another key question in the interview was: what is hardest to replicate in cross-border payments?
Sean Li's answer was clear: not an individual feature, but internally built, globally licensed financial infrastructure combined with integrated products and deployable AI capabilities.
He said Airwallex had spent 10 years establishing partnerships with more than 50 banks and financial institutions. With 87 licenses and collection and payout capabilities in more than 200 countries and regions, he said, it had woven money flows, compliance flows and data flows into one unified underlying network.
That illustrates something: the real moat for future fintech companies will not simply be a better feature, but whether they can compress global complexity into a simple set of interfaces for the user.
Users see an API, a dashboard, an AI assistant and a subscription plan that can be configured in minutes. Those experiences work not because the interface is attractive, but because the network behind it has already been built.
That is also why, in this account, building in-house with open source may have less impact on these platforms than many imagine. Sean Li was direct: large companies certainly can build their own systems, but payments are not merely a matter of technical routing. Open source can help with orchestration and connectivity, but rebuilding globally licensed financial infrastructure is difficult. For such businesses, Airwallex is therefore more likely to supply the underlying cross-border payments and foreign-exchange capabilities.
Put plainly, you can build the orchestration layer yourself, but may not be able to afford to build the underlying supply.
The Future Is Not Finance Being Replaced by an Agent, but Finance Gaining an AI Team
Another question in the interview deserves exploration: will Agents take over corporate finance?
Sean Li's judgment was no. AI Agents directly handling payment decisions are an irreversible trend in his view, but corporate finance will not be taken over entirely. It will enter a new phase of “human-machine collaboration.” An Agent is more like a copilot and early-warning radar, handling repetitive data processing, initial screening and report generation so finance professionals can move from “recorders of the past” to “forecasters of future risks.”
I think that is a grounded judgment. Discussions of Agents today easily go to two extremes: one assumes people will soon be useless; the other thinks an Agent is just a chat wrapper.
The more realistic situation is that Agents first take over processes, not responsibility.
They begin with repetitive work and standardized tasks, then preliminary processing, assessments and recommendations, gradually receiving greater permissions in scenarios that are sufficiently secure and compliant.
Sean Li said Airwallex was working on T0 finance, with the aim of having AI give entrepreneurs a part-time CFO from day one to undertake some internal finance work: reviewing cash flow, forecasting, checking expenses and initiating payments, while submitting decisions for approval when necessary.
This is no longer simply “financial software with an AI assistant.” It is moving toward an AI CFO. It will not replace the person who makes the final call, but will increasingly resemble a finance team that is always online and always watching every data flow.
Airwallex's Ambition Has Long Gone Beyond “Faster, Cheaper Transfers”
If the whole interview were distilled into one business judgment, it would be this: Airwallex has moved beyond a simple payment tool toward being the financial foundation for AI companies operating globally.
Over the next 12 months, the interview says, the AI vertical is the incremental opportunity they are most optimistic about. They hope to make Airwallex standard financial infrastructure for AI companies. For new use cases, they see opportunity in moving corporate financial operations “from recording to action,” using a team of specialized AI agents for highly repetitive tasks such as fund consolidation, payment approvals, expense checks and risk monitoring.
Their forecast for the end of 2026 is also noteworthy: global business collaboration and cross-border payments would evolve from using fragmented tools across multiple PSP providers and banks toward one integrated AI-native financial operating system, allowing finance and business agents to automatically coordinate contracts, reconciliation, payments and capital allocation.
That statement describes more than Airwallex. It also describes an era-wide change in business infrastructure.
Earlier software put “information online.” Later, SaaS made “processes systematic.” AI-native systems are beginning to make “actions automatic.”
When actions become automated, a financial system is no longer simply a bookkeeping, payments or risk-control system. It becomes an operating system that continuously executes, judges and coordinates.
In other words, future competition between companies will concern not only product capabilities, but also “who first has an underlying system that lets agents act safely.”
One Last Word
If you still think of “cross-border payments” as something to add after a company launching overseas goes live, you may already be a little late.
For AI companies, payments, billing, risk controls, compliance and financial automation are no longer back-office issues to handle only after the business grows. They increasingly resemble front-line capabilities that must be considered alongside the product as it goes global from day one.
The AI company that moves fastest in the future will not necessarily have the strongest model. More likely, it will be the one that can not only generate content, acquire users and improve retention, but also collect money around the world, manage it clearly and move it safely.
That is the real hard battle in taking AI businesses global.
Selected Q&A
Q1: What problem does Airwallex actually want to solve?
We began by solving the single problem of cross-border payments, then developed toward addressing the entire financial chain of globally operating businesses: multi-currency collection, treasury management, payouts, subscription billing, reconciliation and, in the future, AI-driven financial automation.
Q2: Why do AI companies need these capabilities more than traditional businesses?
Because AI companies are often inherently global, use more flexible charging models and have leaner organizations that move faster. They encounter local acquiring, multi-entity operations, compute procurement and global payments early on.
Q3: Which parts has AI made noticeably faster?
Account opening and KYC review, payment routing and risk-control decisions, and customer service and product guidance have all accelerated, according to the interview. It cited an approximately 40% improvement in account-opening and onboarding turnaround, and said AI could choose routes, trigger 3DS and score fraud in milliseconds.
Q4: Why, then, say AI has also made things more demanding?
Because the hardest work has shifted from execution to design and governance. Which scenarios permit automatic clearance, which require human approval, how to roll back wrong decisions, how to preserve audit trails, and how to consistently model end-to-end data and expose it through an API—all have become more complex.
Q5: Will Agents completely replace corporate finance?
No. Sean Li's judgment is that Agents will increasingly handle standardized processes within payment and financial decisions, but the ultimate outcome will be “human-machine collaboration.” An Agent is more like a copilot and early-warning radar than a complete replacement for people.
Q6: If a large company builds its own open-source orchestration, does Airwallex still have value?
Yes, in the interview's account. The orchestration layer can be built in-house, but globally licensed financial infrastructure is hard to rebuild. Open source can provide connectivity and orchestration, yet it is difficult to reproduce the underlying license network, bank partnerships, local clearing and compliance capabilities.
Q7: What incremental opportunity is Airwallex most optimistic about next?
The AI vertical. It hopes to make Airwallex standard financial infrastructure for AI companies and continue advancing financial-automation use cases driven by AI agents.