Why we built payouts on top of intelligence
Most payout platforms move the money and check it somewhere nearby. We built Anton the other way round. Here's what that ordering changes, and why I think the rest of the industry will end up there too.

Ask any payout platform whether it does compliance and the answer will be yes. Ask it to draw where the compliance sits, and you'll usually get the same picture: the money in the middle, moving from left to right, and a row of boxes hanging off it. Sanctions screening here. Fraud scoring there. A case management tool underneath, where people write down what the other boxes said.
I've spent twenty years in cross-border payments, most of it in the corners where compliance is never optional, and I've drawn that picture myself more than once. It took me a long time to see what was wrong with it. The boxes aren't the problem. Most of them are very good at what they do. The problem is that the decision, the single most important thing a payout platform does, doesn't live anywhere in that picture. It lives in the gaps between the boxes, and usually in somebody's head.
Anton exists because I stopped believing that picture could be fixed by adding another box.
How payout stacks actually grow up
Nobody sets out to build a fragmented system. It happens in a perfectly reasonable order.
You start with the thing customers pay for: moving money from a business to the people it owes. You find banking partners, you connect rails, you get money landing in local currencies. That's hard, and it's the part everyone sees.
Then a partner bank asks how you know who you're paying. So you integrate a sanctions screening provider and run names against the lists before payouts go out. Reasonable.
Then fraud arrives, because it always does. Accounts are taken over, mule networks find you, someone discovers that a new seller can be paid before anyone notices they're not real. So you add a fraud tool, with its own device signals, its own risk score and its own idea of who a "user" is. Reasonable.
Then a payout is held, a merchant calls, and nobody can explain why. The sanctions tool says one thing, the fraud score says another, and the analyst who made the call has gone home. So you add case management and a policy that says every hold must be documented. Reasonable.
Each step is defensible, and you can pass an audit with all of them. But step back and look at what you've built. Four systems, each with its own data and its own model of the world, none of which knows what the others concluded. And in the middle of it, a person with a lot of browser tabs open, stitching together an answer to the only question that matters: should this money move?
Silje, who works on compliance and operations with us, has a line for this that I haven't been able to stop repeating. In most programs, the analyst is the integration layer. She lived it in iGaming and crypto, and she's right. When the person is the integration layer, the quality of every decision depends on how much context that person happens to be holding at that moment, and all of it walks out the door at the end of their shift.
What changes when the intelligence comes first
At Anton, the intelligence isn't next to the payment. It's the payment's first stop.
When a business sends a payout through Anton, the first thing that happens isn't the money moving. Anton reads the payment. Who is being paid? Is this the same person we've seen before, possibly under a different name, on a different platform, with the same bank account? Where are they, and does that fit everything else we know about them? Has anything changed recently, like new bank details, a new device or a sudden burst of requests? Does the payment make sense for this business and this payee?
Only after that does the money move, or wait for a person to look at it, with the reasons in front of them.
That ordering sounds like an implementation detail. It isn't. It changes at least four things about what a payout platform can be.
The decision has one home. There is one place where the call is made, one record of what was known when it was made, and one explanation. Nobody reconstructs a decision after the fact from four systems' logs. If a regulator, a partner bank or the merchant asks why, the answer already exists, written at the moment it happened, in words a person can read.
The reasons travel with the payment. When the same brain that moves the money also decides whether it should move, the explanation isn't a separate artifact somebody produces later. A held payout carries its reasons with it: what was unusual, what was checked, what a reviewer should look at first. This matters more than it sounds. Explainability isn't a reporting feature. It's what turns a black-box hold into a decision a merchant can trust and a payee can understand.
The network becomes the advantage. A bolted-on check only ever sees the payment in front of it. It's structurally blind to everything outside the platform it's installed on. An intelligence that sits in the middle of every payment, across every business on the network, sees things no single merchant can: the same person turning up under three names on three marketplaces, or the same bank account behind accounts opened in the same week. It also sees the reverse, which gets talked about far less: a payee who has been paid cleanly, everywhere, for a long time.
Good payments get faster, not slower. This is the part people don't expect. The usual assumption is that more intelligence means more friction. In practice it's the opposite. A system that only sees the current payment has to treat everyone like a stranger, so it applies the same suspicion to the seller with years of clean history as to the account opened yesterday. When you actually know who you're paying, you can stop doing that. The honest seller gets paid on time and the reviewer's attention goes to the few payments that deserve it.
A payout, walked through
Here's what that looks like in practice. It's an illustration, not a real case.
A marketplace sends a batch of weekly payouts. Most of the payees are people Anton has seen many times: same names, same bank accounts, same countries, steady amounts. They go straight through, in their local currencies, the way they expect to be paid. Nobody looks at them, and nobody should have to.
One payee is new to this marketplace. On its own that means nothing, because new sellers join every day. But the bank account they've given is one Anton has already seen, on two other platforms, under two slightly different names, all three accounts opened within the same few weeks. Each platform sees an ordinary new seller. Anton sees one person, moving quickly, with no history anywhere.
That payout doesn't move. It waits for a person, and the reviewer doesn't get a score and a shrug. They see the linked accounts, the timing and what's unusual, in plain language, in one place. They might release it within minutes, because there's a perfectly good explanation. They might not. Either way, the decision is theirs, made with the full picture, and the reason is recorded.
Meanwhile every other payout in the batch has already landed.
Strip the intelligence out of that story and you get one of two outcomes. Either the payout goes out, because no single platform could see the pattern. Or the platform, burned before, holds every new seller for manual review, and a hundred honest people wait in line because of one bad actor. Neither is acceptable to me, and I don't think either should be acceptable to the businesses we serve.
Isn't this what everyone says now?
It's a fair question. Every vendor in this industry is "AI-native" this year. Every dashboard glows. I've written before that half the market wants to look like it has better compliance more than it wants to have it.
So here's the test I'd use, on us or on anyone else. Don't ask whether a platform uses AI. Ask where the decision lives.
If the thing that decides whether money moves is a separate product you call out to, with its own data and its own idea of who the payee is, then it's a check. It might be an excellent check, and I've used plenty of them. But it's still a box hanging off the side of the money, and it still can't see what the other boxes know.
If the decision is made at the centre, with everything known about the payee across the network in front of it, and the reasons are written down as it happens, then you have something different.
At Anton there was never a version of the product without the intelligence in it. We didn't add it once the payouts worked. We started there and built payouts around it. That's what we mean when we say Anton isn't using AI. Anton is the AI.
What this doesn't mean
I want to be careful here, because this industry has a habit of overpromising.
Intelligence at the centre doesn't mean nobody reviews anything. It means people review the right things, with the full picture, instead of everything with half of it. Judgment stays human, and the stitching doesn't.
It doesn't mean regulation becomes someone else's problem. Sanctions screening, knowing your customer and monitoring transactions are obligations, and they stay obligations. The difference is that they run as part of one decision instead of four separate ones, and the record of that decision is complete.
And it doesn't mean we're finished. A held payout should never feel like a black hole. The merchant should know why, the payee should know what happens next, and the reviewer should have everything in front of them. That's the standard. We're not done, but that is the standard.
Why it matters to the person getting paid
It's easy to write about all of this as if it's only about stopping bad actors. It isn't, and I don't think that's even the most important part.
Somewhere today a seller on a marketplace checked their phone, saw the money and went back to work without thinking about us at all. A driver finished a shift and the week's pay was already home. A freelancer in another country got paid in the currency she actually spends, the day the work was approved.
The intelligence is what makes those moments possible at scale. It lets us say yes quickly to the people we know, and "not yet" to the few payments that deserve a closer look, without making everyone else wait. Move money quickly without moving it blindly. That's the whole idea, and in my experience it only works if the intelligence comes first.
Ryan Olson is the founder and CEO of Anton.
Ryan Olson
Ryan Olson
