Ellipe 3 — Verifiable Control Execution

You carry the liability. They run the systems.

You signed for a control you cannot watch execute. Someone else runs it, on data you will never hold, inside systems your auditors will never log into — and when the examiner arrives, the question lands on your desk, not theirs.

Three things are true at once, and all three hold wherever this problem is worth solving. You cannot inspect their systems. They cannot hand you the data — it is sensitive, or too large, and receiving it would make the exposure yours. And you cannot simply take their word for it, because their word is precisely what is in question.

This is not a diligence failure, and it is not a trust problem between decent people. Everyone in the chain is acting in good faith with the tools available. There is simply no mechanism — so the gap gets papered over with attestations, and everybody signs, and everybody hopes.

And when hoping fails, the name on the order is yours.

The order has the bank’s name on it

Blue Ridge, January 2024. Evolve, June 2024. Axiom, October 2024. In each case the failure ran through a partner program and the named party was the chartered bank. Blue Ridge then needed regulatory non-objection before onboarding anyone new, and offboarded more than a dozen existing partners to get out from under it.

Regulators have been explicit that banks can no longer rely on partner representations. Which leaves you holding a duty you have no instrument to discharge — and paying for the attempt twice over: once in the compliance function, and again in the hours it takes off the people who should be running the business.

Sources: American Banker / OCC enforcement counts (2024); Bank Policy Institute (2023); LexisNexis Risk Solutions True Cost of Financial Crime Compliance, US & Canada. Figures are industry-wide context, not a projection for your institution.

The spend is not the scandal. The scandal is that it buys you a sample.

Six categories, six blind spots

They also share one blind spot — and it is where the real failures happen.

Controls rarely fail. Populations do.

Enforcement actions overwhelmingly turn on records that never reached the control at all. Your partner is not lying to you; their control ran correctly — on the wrong population, with nothing in the chain capable of noticing. Sampling cannot catch it either: a sample drawn from the included population will never reveal the segment that was excluded before the control began.

So any honest answer has to establish four things at once. Here they are.

Four claims, or it proves nothing

Four steps establish all four of them, and they run in this order.

Four steps, one file at the end

That file is the entire product. This is what one looks like.

Everything, every period, for less than a sample costs

The economics of assurance have always run the wrong way: full coverage was the expensive option, so the industry bought samples and called it evidence. A proof over the whole population costs about what a proof over a sample costs — which quietly inverts the trade-off you have been making for decades.

We are deliberately not putting a savings percentage on this page. Nobody has run it in production yet, and a number we cannot evidence would undercut the one thing we are selling. Quantifying it honestly is a stated goal of the first pilots.

And the same arithmetic holds wherever a control meets a population.

One mechanism. Different words for it.

Every row below is the same sentence — a control, a population, a relying party — in a different industry's vocabulary. We start with FBO ledger reconciliation, where failure is unambiguous. We are not building all ten; we are building the first properly, in a way that does not have to be rebuilt to reach the others.

And none of it asks you to replace anything you already run.

What we do not claim

Stating limits plainly is part of the product. An assurance system whose vendor overstates it is worth nothing to the person who has to rely on it — so this comes before the ask, not after.

Including the largest limit of all — which is simply where we are today.

One gate decides everything