Azentio Software

A broad enterprise software suite with machine learning added to two of its product lines. Strip every model and the core banking platform, the Islamic banking engine, treasury, trade finance, wealth management, insurance policy administration and the whole enterprise resource planning business all continue, because these are transaction processing and record keeping systems whose value is correctness rather than inference.

The financial crime platform states the ordering itself: more than four hundred ready to use rules carry the detection, with built in models helping to spot patterns and connections and cut false alerts. That is models assisting a rules engine, the same construction seen on this roster at Appian and stated more bluntly at Veefin.

Built at C rather than rejected because the models sit inside the regulated operation rather than around it, which is the standing line that separated mycomplianceoffice, red-oak and sybrin from 10x-banking.

Human decision points exist structurally in the products rather than being asserted as a principle. The financial crime platform runs a full investigation and reporting workflow, which places analysts between an alert and any regulatory filing, and the models are positioned as reducing the volume of alerts an investigator must work rather than as disposing of them.

The lending platform is described by a customer in terms of users and approvers, implying an approval chain in the origination path. What is missing is every specific: no confidence threshold, no escalation route, no stated proportion of alerts closed automatically versus routed to a person, and nothing describing what happens at an automated credit decline or an automated screening hit. The false positive reduction claim also raises a question the material does not answer, namely what happens to true positives suppressed alongside the noise.

One figure is published and it will not carry the axis. The financial crime platform is claimed to cut false positives by up to forty percent, which is a bounded marketing figure rather than a measurement: no baseline is stated, no methodology, no institution, no time period, and the qualifier up to means the number is a ceiling rather than a result.

Beyond it there is no validation methodology, no accuracy or recall figure, no drift monitoring, no versioning policy and no independent assessment for either the screening models or the identity verification models. A false positive reduction claim with no accompanying false negative figure is the one number in anti money laundering that cannot be read on its own, because suppressing alerts is trivially easy and suppressing only the wrong ones is the entire problem.

Named institutional customers with qualitative endorsements, and exceptional independent recognition, but no figure attached to any name. Boubyan Bank is quoted on its iMAL upgrade describing substantial efficiency gains and the resolution of key operational challenges, Tamam Financing on a lending deployment describing smoother processing and improved comfort for users and approvers, and Philippine National Bank is named on core and digital modernisation.

A further anti money laundering deployment reducing manual effort is attributed only to India's leading fintech, unnamed. The strongest quantified claim, a reduction in false positives of up to forty percent, is a product claim by the vendor rather than a result reported by an institution.

Held at B on the same basis applied to Opensee and Aurionpro in this session: analyst evaluations are assessors rather than parties with money at stake, and an A requires a number attached to a named institution.

Nothing states whether customer data trains the vendor's models. The question has real weight here because the financial crime platform runs across many institutions in overlapping markets, and pattern detection benefits from pooled typologies in a way that would be valuable to the vendor and invisible to any single bank. No exclusion of client data from a training corpus, no separation commitment between institutions, and no statement of controls around the models are published. Against the reference set of Mortgage Capital Trading, Needl and AlphaSense, all of which answer this plainly, the silence is a choice.

No privacy posture is published. Nothing addresses retention, purge on termination, or the handling of the identity documents and biometric data that the digital onboarding capability necessarily collects. The biometric point is the sharper one: face matching and liveness detection generate facial templates, which several jurisdictions in this vendor's markets treat as a special category of personal data with distinct consent and retention obligations, and none of that is addressed anywhere.

No security certification is enumerated in the material reviewed. No SOC report of either type, no ISO 27001, no penetration testing statement and no trust portal was located, which is a striking absence for a supplier running core banking, holding customer accounts and processing identity documents and biometric data for banks across three continents.

Queued check: enterprise vendors of this size commonly hold attestations behind a request process rather than publishing them, and a private equity owned group selling core banking into regulated institutions would ordinarily be required to produce them during procurement.

A software supplier with no licence, authorisation or supervisory relationship of its own. One credential is genuinely uncommon and is recorded here without carrying the grade: certification from the Accounting and Auditing Organization for Islamic Financial Institutions covering the Shariah compliant product set and profit calculation engine.

That is a real assessment by a recognised standards body against published standards, and it is stronger than most badges in this index, but under the standing bar a conformity assessment does not read across as regulatory standing. No enrolment, sandbox participation or registration was found. The company does publish interpretation of central bank artificial intelligence guidance for its markets, which is category commentary rather than a position of its own.

No fairness position, disparate impact testing, protected characteristic handling, model inventory or named governance framework is published. Two concrete exposures sit unaddressed. Face matching and liveness detection are deployed for account opening across Asian, Middle Eastern and African markets with no demographic performance breakdown and no false reject rate by cohort, which is the documented weak point of facial recognition and the same gap recorded against other proofing vendors in this index. Separately, anti money laundering screening and risk scoring determine which customers are subjected to enhanced scrutiny or refused onboarding, and no fairness position accompanies either.

No recourse position is published. Nothing states what happens when a screening model produces a false match against a person, when liveness detection wrongly rejects a legitimate applicant, or when a credit decision is wrong, and nothing addresses whether the affected person is told automated processing was involved.

The pattern follows the standing finding for software suppliers: the bank is the regulated party answerable to its supervisor and its customer, while the supplier that built the screening and biometric models sits outside the conduct perimeter. The person wrongly matched or wrongly rejected has no relationship with the vendor at all.

No model, family, version or provider is named for any capability. The financial crime models, the identity verification and face matching, and the liveness detection are all described by function and none by dependency. Biometric matching in particular is a capability most vendors license rather than build, and this index has repeatedly found the underlying supplier named by the supplier rather than by the vendor, so the standing sourcing tell applies directly here on a later pass. No statement exists on whether models are built in house or licensed, no version policy and no per capability breakdown.

The vendor supplies the core itself, which is the deepest position available on this axis. The iMAL platform is the system of record for the banks that run it, and the surrounding modules for lending, treasury, trade finance and wealth attach to it natively rather than through connectors.

The Islamic banking engine goes further than integration into domain mechanics, automating profit accruals, allocations, reserves and distributions and managing profit sharing pools, which requires the accounting core rather than sitting beside it. The digital layer is documented as deployable both onto the vendor's own core and onto any third party core, and the architecture is described as API first and open banking ready.

Held at A on the strength of owning the core, with the qualification that no specific third party core banking platform is named as a certified integration, so the interoperability claim is generic where the native depth is not.

Architecture is described as cloud ready and API first, which states a design property rather than a deployment position. No cloud provider is named, no regions are listed, no residency commitment is given, and there is no statement of whether deployments are on premise, in a managed cloud or hybrid, which is a live question for core banking in particular.

The gap is material for this buyer base: banks across the Gulf, South East Asia and Africa operate under national data localisation regimes that make processing location a first order procurement condition rather than a technical detail.

No pricing information of any kind was found across core banking, lending, financial crime or enterprise resource planning. No licence basis, no per institution or per transaction metering, no module pricing, no implementation estimate and no indicative contract size. Every route into the products is an enquiry or demonstration request. For a supplier selling core banking replacements, where total cost of ownership over a decade is the central purchasing question, the absence is total.

A broad enterprise software suite with machine learning added to two of its product lines. Strip every model and the core banking platform, the Islamic banking engine, treasury, trade finance, wealth management, insurance policy administration and the whole enterprise resource planning business all continue, because these are transaction processing and record keeping systems whose value is correctness rather than inference.

The financial crime platform states the ordering itself: more than four hundred ready to use rules carry the detection, with built in models helping to spot patterns and connections and cut false alerts. That is models assisting a rules engine, the same construction seen on this roster at Appian and stated more bluntly at Veefin.

Built at C rather than rejected because the models sit inside the regulated operation rather than around it, which is the standing line that separated the compliance and proofing platforms already indexed from the core banking vendor rejected on this test.