Oscilar

Models do real and named work here rather than serving as a label, including credit models for balance prediction and spending power, repayment prediction, transaction categorisation and cash flow scoring, plus agents that triage alerts and draft investigation narratives. The platform sits on a no code decisioning substrate with an integration hub of more than eighty data sources, and that substrate is what a risk team operates day to day when composing policies.

Apply the removal test and a capable orchestration and rules product remains, which places this with the decisioning platforms rather than the model native vendors despite the company positioning itself as the originator of an AI category.

Automation is extensive and the controls around it are stated and mechanical rather than aspirational. Policies are authored by the risk team through a visual builder or in natural language, and crucially they can be tested and validated before deployment, so a new credit or fraud policy is proven against expected outcomes rather than discovered in production. Real time performance monitoring lets a team see how live policies behave without writing queries. The company states that human guided governance ensures compliant decisions, and analyst feedback is designed into the loop so reviewers correct the system rather than merely consuming it.

Naming the individual credit models rather than referring to proprietary AI is itself a disclosure most competitors do not make, and it gives a validator something specific to interrogate. Policy validation before deployment provides pre production evidence, and real time performance analytics provide the ongoing monitoring that supervisory guidance expects, both delivered as product rather than promised in a document.

The formal package remains absent: no model documentation, no validation summary, no published accuracy or error rates for the named models, no retraining or drift monitoring cadence and no stated support for a customer's own validation.

The named customer list is long and spans institution types, including a large consumer lender, a global money transfer business, a payment processor, a community bank and a range of card, fleet and lending fintechs, which is unusual disclosure for a company of this age. A partnership with the money transfer customer is described as reaching more than 200 countries and territories.

Outcome figures are specific, citing 45 percent fewer false positives, five times faster policy deployment, three times faster case resolution and decisions under 100 milliseconds. What keeps this off the top grade is attribution: those figures are presented as platform aggregates rather than as a named institution's measured result.

The agents are described as trained on the institution's own standard operating procedures rather than generic templates, and analyst feedback is said to refine recommendations over time, which implies learning scoped to the customer rather than pooled across the base. That is a better default than the shared learning several vendors in this index describe without bounding, though the boundary is implied rather than stated. Published subprocessor disclosure supports it. Not addressed: which model providers sit behind the agents, whether any signal crosses between customers, and how agent output is evaluated for accuracy before it reaches an analyst.

Oscilar publishes a dedicated subprocessor list alongside its privacy policy, cookie policy and security page, and that single artifact puts it ahead of nearly everything else assessed in this index. A subprocessor list tells a bank privacy office which third parties can touch its data, which is the first question in any vendor privacy review and the one almost no vendor here answers publicly.

It matters more given the platform routes applicant and transaction data through more than eighty connected data sources. What is still absent is a stated position on financial privacy service provider obligations and a published retention schedule.

A dedicated security page and a separate subprocessor page are both published and linked from site navigation, which is a navigable assurance surface rather than a single line of reassurance, and subprocessor disclosure in particular is a specific and checkable commitment that most vendors in this index avoid making.

What was not located is the certification detail itself: no enumerated framework list, no attestation scope and no audit period, so a buyer can see who touches the data without being able to confirm which controls have been independently examined.

Oscilar supplies technology and holds no licence, the expected posture, and it holds one formal designation worth noting: preferred partner status with the body that governs the automated clearing house network, covering account validation and fraud monitoring, which is an admission process rather than a marketing badge.

Product scope maps onto bank secrecy and anti money laundering obligations including sanctions screening and suspicious activity reporting, with dedicated treatment for sponsor banks whose partner oversight duties are under active supervisory pressure, and published material engages with European operational resilience and capital frameworks.

Credit underwriting is a headline product for both consumer and commercial lending, which places these models inside equal credit opportunity rules where adverse action reasons must be specific and disparate impact is a live supervisory concern, and no fair lending testing, demographic performance analysis, adverse action reason code documentation or independent audit is published.

The named credit models sharpen the issue rather than softening it: balance prediction, repayment prediction, transaction categorisation and cash flow scoring all derive from banking transaction data, which regulators have repeatedly flagged as correlating with income volatility, employment sector and geography, so the fair lending exposure is concentrated exactly where the modelling is most sophisticated.

Policy validation before deployment and real time performance monitoring give an institution the means to catch a bad rule before it reaches applicants and to see drift after it does, which places the customer in a position to be accountable. The vendor offers no commitment of its own: no accuracy guarantee for the named credit models, no remediation term, and no correction route for a borrower declined on a cash flow or repayment prediction they will never see.

Oscilar publishes a dedicated subprocessor page, an artifact almost absent from this index and the one that actually answers the fourth party question a bank vendor review asks first. Alongside it sit a security page, an integration hub of more than eighty named data sources with a published marketplace, and named proprietary credit models rather than an undifferentiated reference to AI. A buyer can therefore enumerate most of the chain. The generative and agentic model providers remain unnamed.

More than eighty data sources connect through an integration hub with one click connections and a published marketplace, spanning identity, bank account, credit and compliance data so a workflow can assemble a full risk view without bespoke engineering per provider. Account validation through the clearing house network partnership plugs the platform into payment rails directly rather than through an intermediary. The build surface is dual, with no code composition for risk analysts and natural language generation of workflows, so a rule change does not require an engineering release cycle.

Delivery is cloud hosted software as a service, with a customer partnership described as spanning more than 200 countries and territories, which necessarily implies multi region operation and cross border movement of applicant and transaction data. The subprocessor disclosure tells a buyer which parties are involved, which is more than most vendors offer, but it is not a residency statement. No hosting regions, in country residency options, transfer mechanisms or tenancy separation details were located in this pass.

No rates, tiers, billing unit or minimum are published and paths lead to a demo request. The omission sits awkwardly against the platform's central commercial argument, which is consolidation: it claims to replace separate fraud, compliance and credit tools with one system, and a buyer cannot weigh that against the several licences being displaced without a number on either side.

Six institution types each carry dedicated material: banks, credit unions, sponsor banks, fintechs, payments companies and digital asset platforms. Functional coverage is equally wide, running the whole customer lifecycle from consumer, business and merchant onboarding through fraud, scams, transaction monitoring and sanctions screening to consumer and commercial credit underwriting, portfolio monitoring and collections.

Fraud typologies are named at a level of specificity most vendors avoid, including authorised push payment fraud, investment manipulation scams and romance scams, which indicates the detection logic was written against real typologies rather than categories.