The cost nobody talks about
Ask a financial institution what their AML and fraud compliance costs them, and they'll quote you headcount. Analysts. Case managers. A MLRO and their team.
What they rarely account for is everything else.
The data engineers whose job it is to maintain and retune a monitoring system that was never quite right out of the box. The vendor consultants who show up every quarter for a "model optimization" engagement. The professional services hours billed every time a regulatory requirement shifts and the thresholds need adjusting. The implementation cycles every time a new product line needs to be covered.
Compliance technology has built an entire economy around the premise that the software needs permanent human support to function. And financial institutions have accepted it — not because they believe it's right, but because they've never seen a credible alternative.
Meanwhile, the compliance officer — the person with the deepest understanding of the institution's risk profile, the regulatory context, and what "suspicious" actually means in their specific business — is not in the driver's seat. They're in the back, waiting for someone to finish a sprint before they can change a rule.
That's what Pascal and I set out to fix when we started building Marble in 2023.
A swarm of agents, one supervisor
Marble V1 will build around a different model: a swarm of AI agents operating under the direct supervision of the compliance officer — not the vendor, not the engineering team, not an external consultant.
These agents will handle the work that has historically required human intermediaries: tuning detection rules, calibrating customer scoring models, adjusting thresholds as risk patterns evolve, gathering investigation context, pre-qualifying alerts, mapping behavioral patterns across a customer's full history.
They do it continuously. Autonomously. And they surface their work to the compliance officer in a way that keeps her in control of every consequential decision.
This is the distinction that matters. Marble isn't replacing the compliance officer's judgment : it's eliminating everything that was getting in the way of her using it. When an alert fires, she isn't starting from scratch, waiting for a data pull or a vendor report. She has everything she needs to answer the one question that is genuinely hers: does this make sense for this specific customer?
Why now
V1 is not an arbitrary version number. It marks the first time everything is truly in place.
Marble now covers the full compliance journey seamlessly, from screening and transaction monitoring, connected to the best available data sources, through to automated detection and AI-assisted investigation. No gaps. No brittle integrations between modules that were built independently. One platform, working end to end, the way it was always supposed to.
That matters for our customers, who can now deploy a single coherent system rather than stitching together point solutions. And it matters for us : it means we can build forward from a complete foundation rather than filling holes.
This is the milestone V1 represents. It’s not perfection but completeness.
Why Marble, specifically
We're at a moment where, frankly, anyone could probably vibe-code something that looks like a reliable AML and fraud platform. The models are powerful. The APIs are accessible. The demo would probably be convincing.
But no one wants to bet their business on it.
What Marble offers isn't just software - though that's at the core. It's years of experience actually operating it - in production, in regulated environments, across institutions that cannot afford to get this wrong. We've seen how detection models degrade over time. We know what regulators ask for and what they're skeptical of. We've built through regulatory cycles, through fraud pattern shifts, through the rapid evolution of AI capabilities.
And we've made a commitment that comes with the product: continued investment to keep Marble state of the art, now and in the years to come. Core updates every 2 weeks. The regulatory landscape will keep moving. Fraud typologies will evolve. AI will advance fast than anyone can fully predict. We've built a team and a business model designed to stay ahead of all of it — so our customers don't have to hire someone to do it for them.
This is the difference between a vendor and a long-term partner.
What this means in practice
The structural cost saving from Marble isn't primarily about analyst hours — though that's real. It's about eliminating the hidden infrastructure of human support that the industry has normalized.
When the system will tunes itself, you don't need a retainer to do it for you. When agents surface investigation context automatically, you don't need bespoke data pipelines for every workflow. When the compliance officer can directly supervise and improve the system, you don't need technical intermediaries translating between her needs and the software.
You get effective detection — not headcount-dependent detection.
Available on-premises or SaaS. With an open-source core, because compliance infrastructure must be fully auditable - by your IT security team, by your regulators, by anyone who needs to understand how decisions are being made.
Where we are
More than 100 financial institutions - from the United States to Singapore - are already running on Marble, with a strong presence in Europe, where the regulatory bar is among the highest in the world.
That's not a marketing claim. It's evidence that the model works, in the environments where it's hardest to make it work.
What comes next
V1 is the foundation. It is not the ceiling.
The institution that runs on Marble doesn't accept that compliance effectiveness is constrained by vendor calendars or engineering bandwidth. They have a compliance officer who is genuinely in control - with agents working under her supervision, handling everything that doesn't require his or her judgment, so he can focus on what does.
That was the vision. This is what it looks like in production.
Marble is an autonomous AML and fraud monitoring platform — screening, transaction monitoring, and customer scoring in real time. Open-source core. On-prem and SaaS. checkmarble.com
