This post is inspired by Marc Boiron’s article, “The real bottleneck in finance is settlement — not payments,” published in Fintech Weekly.
A decade ago, “fast payments” meant the app didn’t crash. The confirmation screen showed up instantly. The money arrived later — sometimes hours later, sometimes the next day. We collectively called it “instant” because the notification was fast, even if finality wasn’t. Boiron’s piece captures that gap well: the industry got very good at making money movement feel real-time, while the underlying settlement reality stayed stubbornly asynchronous.
That polite fiction is breaking down — not because consumers suddenly got picky, but because finance is becoming continuous and software-driven. Treasury systems rebalance automatically. Marketplaces trigger payouts through APIs. Risk engines move funds based on rules, not office hours. And every one of those systems needs a simple thing legacy rails still struggle to provide:
A clear, timely answer to: “Did it actually settle — and is our ledger right?”
Payments have gotten faster. Settlement and back-office truth haven’t.
What “settlement” feels like on the ground
If you’ve never lived in FinOps, settlement sounds abstract — like something that happens “in the plumbing.” In reality, it’s painfully concrete. It looks like: a finance lead opening three portals (processor, sponsor bank, internal ledger) and realizing yesterday’s totals don’t tie. It’s not a “big” mismatch — maybe it’s 0.3% — but it’s spread across hundreds of items, and now the team has to find the needles in the haystack.
It’s the constant drip of exceptions:
- a return that shows up days later,
- a chargeback that reverses revenue weeks later,
- a payout file where one partner’s IDs don’t match your IDs,
- a “successful” status that later fails at clearing,
- a fee schedule applied slightly differently than you modeled.
And because the business keeps running while the truth lags, your team ends up doing the least scalable work in the company: manual matching, hunting, explaining.
This is why “settlement” becomes the bottleneck once volume and partners ramp up. It’s not that you can’t move money. It’s that your organization can’t confidently answer what happened without a human doing detective work — exactly the operational drag Boiron points to when he argues the bottleneck is settlement, not initiating payment.
The metrics that make this real (and uncomfortable)
At scale, tiny failure rates become an operational machine. Take U.S. securities settlement. Even after moving to T+1, the ecosystem still sees meaningful fail rates: DTCC reported “CNS fails around ~2% and non-CNS fails around ~3% in mid-2024 metrics around the T+1 transition.”
Or look at ACH. The network is enormous — 35.2B payments in 2025 (with Same Day ACH reaching 1.4B payments). Now overlay risk thresholds: Nacha’s rules set an unauthorized return rate threshold at 0.5%. At high volume, “half a percent” is not small. It’s a line item. It’s enforcement. It’s reputational and partner risk.
And then there’s the time tax. Across finance ops literature and vendor benchmarks, manual reconciliation regularly consumes a material chunk of finance team capacity, especially in multi-partner environments where data formats and identifiers don’t line up cleanly. Even if your company is “good,” that’s still days every month spent on work that produces zero customer value — just preventing the business from flying blind.
Why this is risky (not just annoying)
In a fast-moving fintech, the scariest failures aren’t the ones you notice. They’re the ones you discover late, when: the partner says a file was wrong, an auditor asks you to prove completeness, a sponsor bank wants evidence of control, or your CFO realizes cash and revenue were overstated.
In other words: pending isn’t a status. It’s risk. Boiron frames this as the industry’s core problem: we optimized movement, not certainty.
Settlement risk creates a particular kind of fragility:
- You can’t confidently answer “how much cash do we have, really?”
- You can’t trust “unit economics” if fees and reversals arrive later and inconsistently.
- You accumulate operational debt: a growing backlog of unexplained differences that becomes harder to unwind every month.
This is how teams end up with a reconciliation process that only two people understand, running off a brittle set of SQL queries and spreadsheet macros — until it breaks.
Where stablecoins (and AI) change the shape of the problem
A lot of people talk about stablecoins like they’re a “new payment method.” That’s not the interesting part. The interesting part — aligned with Boiron’s argument — is that stablecoins can compress the gap between execution and settlement, and enable 24/7 settlement expectations.
The volume is no longer theoretical, though different sources measure different things (raw transfers vs “real payments”):
- TRM Labs highlights stablecoins’ massive and growing transaction volumes.
- Artemis estimates on-chain stablecoin settlement at very large annualized levels (after de-noising).
- McKinsey argues actual stablecoin payments are smaller than raw transfer totals, but still material and rising.
Put together: stablecoins increase availability of settlement, but they do not eliminate reconciliation. They shift where reconciliation happens and raise the standard for “real-time truth.”
AI accelerates the need. As finance teams adopt agentic workflows — bots that trigger payouts, rebalance liquidity, or execute hedges — the tolerance for ambiguity drops. Machines don’t “feel” uncertainty. They just act on states. If the state is wrong, they’ll automate the mistake at scale.
So the future isn’t “faster payments.” It’s continuous finance that demands continuous truth.
The Rexi POV: Settlement will get faster. Reconciliation must become real-time.
Here’s our point of view at Rexi: Settlement finality will keep compressing (driven by stablecoins, tokenization, and 24/7 expectations), but the real wedge for operational excellence will be the systems that keep your financial source-of-truth correct as complexity grows.
The winners won’t be the teams that “move money” the fastest. They’ll be the teams that can say — at any moment:
- What settled.
- What didn’t.
- Why not.
- What it means financially.
- And what action is required.
Without waiting for end-of-day. Without heroics. Without the spreadsheet gauntlet.
How Rexi solves the pain your FinOps team lives with
Rexi is built for the moment when “payments worked” but finance is drowning anyway. We automate the day-to-day reality: Rexi ingests and normalizes messy data from banks, processors, PSPs, custodians, and internal ledgers — even when formats and IDs don’t line up. We reconcile continuously (not just in batches), so breaks show up when they happen, not at month-end. We detect discrepancies in real time, route them through an exception workflow finance teams can actually operate, and explain breaks with auditable evidence (what changed, which side is missing, what rule failed). And for banks or regulated partners, we support bank-hosted / on-prem deployments when the environment demands it.
The punchline is simple: your team stops doing forensic accounting every day and starts operating the business with confidence.
Why this matters right now
If you’re scaling a fintech, bank program, marketplace, insurer, or payments business, your complexity is compounding in three directions at once: more volume, more partners, more automation. That combination turns settlement ambiguity into an operating constraint — and eventually into an existential risk.
Boiron’s piece calls out the right root cause: we built a world where money can move quickly, but certainty still arrives late.
From where we sit, the practical version is even more direct: The bottleneck is the gap between what your systems say and what actually happened. Closing that gap — continuously — is what unlocks speed safely.