Neobanks are adding crypto rails and most are doing it backwards.

I watched this movie twice. Revolut in 2018. Then Wise in 2024. Same problem, two different scripts.

Revolut launched crypto trading, hit regulators, spent five years arguing. Wise launched stablecoins and it just... worked. Not because their engineers were smarter. Not because they had more money. They just asked the hard question first instead of last.

See, most fintechs build backwards. They nail the product. Then custody. Then they call a lawyer and hope he says yes. Doesn't work that way in crypto. It actually doesn't.

Stop treating compliance like paperwork

Traditional payments let you be sloppy about this. Build something nice. Ship it. Compliance is the thing that happens after product launches, right? ACH and SWIFT already exist. You plug in. They handle the boring bits.

Crypto has no established plumbing. Every transaction lives on-chain. Every wallet is a regulatory identity (or it isn't, depending on who you ask). A token that's fine in Malta might be a security in the US. Compliance isn't a box you check after launch. It's the box that determines whether launch is even possible.

Wise figured this out. They got licensed money transmitter status in every single jurisdiction they wanted to touch. That license defined the walls. Then they built the product inside those walls. Revolut did it backwards. Product first. Discovered walls later. Spent years renegotiating with regulators while their product couldn't expand.

The lesson tastes bitter but it's free. Compliance first actually moves you faster.

Custody is not the starting line

Everyone gets this wrong. Fintech team sees crypto, immediately asks "who holds the money?" Sign Fireblocks contract, announce it on Twitter, declare victory.

Custody solves one thing and one thing only. Where do the assets sit. It doesn't solve whether Singapore lets you trade that asset. It doesn't solve whether stablecoins fall under MiCA. It doesn't solve whether the SEC thinks you're running an unregistered securities exchange.

Custody also doesn't cover your downside. Custody provider gets sanctioned tomorrow? Your users' money is gone and your terms of service are suddenly a lie they can sue over. Your wallet balances drift from your database? Which one's true? You need blockchain-as-source-of-truth architecture and custody providers don't give that to you.

Reconciliation between on-chain and fiat is its own problem. Did the mint happen right? Are gas fees being charged correctly? When the merchant converts back to fiat, does that settlement actually match what you said would happen? This isn't technical debt. This is architectural debt.

Regulatory reporting is your problem, not the custody provider's. Regulators want transaction-level data for every off-chain-to-on-chain movement. You have to build that. They're not reading the blockchain.

Custody is important. Just not first.

Fiat reconciliation will break you if you're sloppy

Traditional fintech is simple. Someone pays. Money moves. System tells them. Balance updates. Done.

Crypto is three movements at once. Buy stablecoin (settlement). Move to wallet (on-chain). Send to merchant (on-chain). Merchant gets it (on-chain). Convert back to fiat (settlement). Bank deposit (settlement). Six transactions, six reconciliation problems.

Wise built reconciliation as architecture. Every corridor has a protocol. Users see real finality times. App explains the delays instead of just showing numbers.

Revolut treated it like a technical glitch they'd fix later. In 2023, users' balances didn't match their actual holdings for days. That's not a bug. That's what happens when you skip the architectural layer.

The actual winning order

You want to know how to not blow this up? Here's the sequence.

Get licensed somewhere that matters. EU means PSD2 or e-money institution license. US means money transmitter licenses in the states you touch. Pick a jurisdiction where the regulator has actually written guidance on crypto, not mystery language.

Build the compliance architecture next. Document your risk appetite. Know your KYC thresholds. Build sanctions screening that actually works. Write the regulatory reporting pipes. Get lawyers to sign off everywhere. This sounds bureaucratic. It's actually the thing that saves you.

Design payment flows around the boundaries you just built. You now know which assets you can actually support. You know which jurisdictions you can operate in. You know what settlement times you can promise users. Design for that reality. Don't fight it.

Then partner for custody. You know exactly what features you need now. You know what custody needs to do. Stop guessing.

Launch in a sandbox or limited scope. Get real data. Talk to regulators. Fix things. Only then expand.

Wise, PayPal, Stripe. These three did this. Revolut, FTX, BlockFi. These three got creative with the order and it went bad. Some of them shut down. Some got cornered by regulators. All of them burned capital fighting battles they could've avoided.

Compliance-first looks slower. It's actually faster. You know where the walls are. You move inside them. Product-first hits the walls at scale and has to redesign everything. One takes six months. The other takes five years.