For many years I’ve been involved in FinTech, leading teams that integrate payment methods. Wire transfers, SEPA, debit cards, credit cards, mobile wallets, PayPal, Alipay. Each one has its own failure modes and each one has kept a tech team busy for months at some point.

ACH was always the one I flinched at. It’s old, it settles in batches, the return window stretches for weeks, and it’s a pull rather than a push. Somebody hands you an account number and you reach into their account for the money. Every time we scoped an ACH integration the same question came up, and it was never really a technical one: is the risk worth the reward?

Yesterday the Federal Reserve launched FedNow, and for the first time that question has a different shape.

ACH earned its keep #

It’s easy to be rude about ACH, and mostly unfair. It has moved the payroll of the United States for fifty years. It handles enormous volume, it degrades gracefully, and it works at banks that have no appetite for changing anything. That’s a real achievement and nothing that came after it has had to prove itself at that scale yet.

The problems are structural rather than sloppy. Batch settlement means you find out about failures long after the fact. The pull mechanism means the risk sits with whoever initiated, so as the merchant you’re exposed to a return you can’t see coming. Unauthorized consumer debits can come back at you up to sixty days later, which means an ACH payment isn’t really final on the day it clears, it’s provisional for two months. If you’re building anything that hands over goods on receipt of funds, that gap is the whole problem.

The developer experience doesn’t help. The NACHA file format is fixed-width records and batch headers, which feels less like an API and more like an afternoon of COBOL.

What FedNow actually changes #

Three things I care about:

It’s a push. The payer’s bank initiates and the funds move. That inverts the fraud model. You’re no longer waiting sixty days to find out whether you kept the money, because there’s no return window to wait through. Instant and irrevocable are the same property viewed from two sides, and it cuts both ways: a mistaken payment is a much harder conversation when there’s no mechanism to pull it back.

It’s ISO 20022. Structured, extensible, and already the standard everywhere else. If you’ve integrated anything in Europe in the last decade you’ve met it. It carries remittance data in the message itself, which is the thing ACH never did well and which everybody worked around with reconciliation glue.

It’s the Fed. This is the part I’d underweighted at first. RTP has existed since 2017 and does real-time interbank payments in the US already, so FedNow isn’t the first instant rail here and anyone telling you otherwise is selling something. What FedNow has that RTP doesn’t is the Federal Reserve’s existing relationship with essentially every depository institution in the country, including the small banks and credit unions that were never going to join a network owned by the largest banks. The interesting number isn’t the settlement speed, it’s the eventual reach.

The part nobody knows yet #

Coverage on day one is 35 institutions out of more than nine thousand. Adoption is the entire story and it’s going to be slow, because the constraint was never the protocol. It’s core banking software at institutions that upgrade on a decade cycle.

SEPA Instant went live in 2017 and took years to become something you could assume was there. My guess is FedNow follows a similar curve, quietly, and one day you notice ACH has become the thing you use for payroll and not much else.

What I’m less sure about is whether irrevocability turns out to be the feature or the bug. Push payments move the fraud from returns to social engineering, and the UK has spent years discovering how expensive that gets. We’ll find out.