Fincom

This is deep technology rather than a model wrapper, and the distinction cuts both ways on this axis. The patented core converts names into mathematical representations of their pronunciation using 48 algorithms drawn from phonetics, computational linguistics and mathematics, which is deterministic transformation rather than inference.

Machine learning is genuinely present, described as purpose-built and supervised, working inside a fuzzy logic engine that weighs attributes such as date of birth, address and identifiers alongside distance scoring on the phonetic match. Held at B because removing the learned layer leaves the phonetic engine, which is the differentiator the patents cover.

The product surfaces alerts for human adjudication rather than making decisions, and it ships a case management system so compliance officers can work them, which is the correct division in a domain where the decision to block a payment or file a report carries statutory consequence for the institution. The stated purpose of the technology is reducing the alert volume officers must clear rather than clearing it for them. What is absent is any description of thresholds, of what proceeds automatically when no alert fires, and of how a bank tunes sensitivity against its own risk appetite.

The published measurement is the most complete in this category and covers both error directions, which is what distinguishes a real claim from a marketing one. Alert rates are stated against a named industry baseline, falling from around 30 percent to under 3 percent, and the company explicitly claims this occurs without missing hits, addressing the recall side that any false positive reduction must answer.

Latency is quantified at under 200 milliseconds against named commercial watchlists. Results are described as validated across numerous United States banks rather than internally derived. The company also sells model validation as a service, which means it operates inside the discipline its buyers must satisfy.

The Series B was led by a major exchange operator's venture arm with a global financial group participating, and came with a global partnership with that exchange's financial crime technology business, which is an incumbent in this exact category choosing to partner rather than compete.

Adoption is described as dozens of United States banks including tier one institutions, financial institutions across the European Union and United Kingdom, international regulators and homeland security agencies. Three named organisations resell or distribute the product, including one of the largest core banking software vendors acting as a worldwide reseller and two global professional services firms. The company appears on an established regulatory technology ranking.

No data boundary statement was located. The exposure is structurally lower than for most vendors here because the core engine is a patented deterministic transformation rather than a system that learns from customer traffic, so a bank's screening volume does not obviously improve a competitor's results.

That advantage is undercut by the supervised learning layer, which by definition trains on labelled outcomes, and nothing states whose adjudications provide those labels or whether tuning is tenant specific.

No data protection agreement, retention schedule, subprocessor list or deletion commitment was located. The platform processes customer names and identity attributes including dates of birth, addresses, contact details and identification numbers, screens them against third party watchlists, and generates alert records that become part of a bank's compliance file. Deployment spans jurisdictions with materially different data protection regimes and none of the handling is described.

No attestation, certification, trust centre or enumerated framework was located. Tier one banks, regulators and homeland security agencies have all completed supplier assessment on this vendor, which is about as demanding a set of reviewers as exists, and none of the resulting assurance is published for a prospective institution to examine.

The entire product is organised around a named regulatory instrument, the United States sanctions administration regime, with screening built to satisfy its requirements and the two dominant commercial watchlist databases named as screening sources. Further named obligations include payee verification, which is a specific European requirement, perpetual customer due diligence, and model validation, the discipline supervisors expect for any system influencing compliance decisions.

International regulators and homeland security agencies are among the customers. A customer specifically describes the vendor supporting it through an audit, which is the practical test of whether a compliance product survives examination.

The fairness effect here is real, measured and largely unremarked by the company itself. Sanctions screening false positives fall disproportionately on people whose names originate in non-Latin scripts, because transliteration from Arabic, Cyrillic or Chinese produces many valid spellings and conventional matching flags them all, so an industry alert rate around 30 percent is not evenly distributed across customers.

Matching on phoneme rather than spelling across 44 languages, and cutting alerts to under 3 percent while claiming no loss of true hits, directly reduces a burden borne mostly by people with non-Western names, whose payments are delayed and whose accounts attract scrutiny for reasons of orthography rather than conduct. Held at B because the company frames this as efficiency rather than equity and publishes no analysis by name origin or population.

No guarantee, indemnity or correction process was located. The institution is well served, with published accuracy figures, validation across peer banks and demonstrated support during examination. The individual is not addressed, though the product improves their position structurally: a person whose payment is held because their name resembles a sanctioned entity is not told why, cannot see the match, and has no route to have a recurring false match suppressed, and the platform's value is precisely that far fewer people end up in that position.

The two dominant commercial sanctions and politically exposed person databases are named as screening sources, which is the dependency that determines whether a screen is valid at all, since list currency governs compliance more than matching quality does. The core engines are the company's own, patented and described in technical detail down to the number and disciplines of the algorithms involved. What is not disclosed is any base model behind the supervised learning layer, any hosting arrangement, or a subprocessor list.

Distribution runs through the institutions banks already buy from: one of the largest core banking software vendors resells the product worldwide and lists it in its application marketplace, two global professional services firms act as channel partners, and a major exchange group's financial crime technology business entered a global partnership. On the data side the platform screens against the two dominant commercial watchlist providers by name.

The engine is described as structure, language and data source agnostic, and it operates across wire, international messaging, automated clearing and instant payment rails, which is the practical requirement for a bank running several at once.

No hosting provider, region selection, residency commitment or private deployment option was located. The gap matters because the customer base spans United States banks, European and British institutions and government agencies, several of which impose strict requirements on where screening data is processed, and because sub 200 millisecond latency in payment flows implies deployment topology decisions that are not described.

No price is published and the saving is quantified consistently and specifically, at more than 80 percent reduction in sanctions screening operational cost, stated elsewhere as over 90 percent on wire and international payment screening, with the company noting the results are validated across numerous United States banks rather than modelled.

For a compliance function whose cost is dominated by staff clearing alerts, a verified reduction of that magnitude is the number that determines the business case. What is charged for it, and whether by volume, institution or rail, is not stated.

Coverage is exceptional on every dimension. Buyers span tier one United States banks, European and British financial institutions, payment service providers, fintechs, international regulators and homeland security agencies. Payment rails covered include wire, international messaging, automated clearing, real time payments and the newer instant settlement networks.

Applications run from sanctions screening and payment screening to payee verification, perpetual customer due diligence, trade finance and model validation. Matching handles individual and corporate names, bank names, vessel and aircraft identifiers, addresses, product names, tariff codes, bank identifiers, free text and identity numbers, across 44 languages in Latin and non-Latin scripts.