ADVANCE.AI

Document authentication, liveness and deepfake detection run on first party models, alongside credit scoring and fraud models. The orchestration layer institutions use to assemble journeys and connect their own data would still sell without them.

Onboarding is built to finish without a person, and control sits at design time, where the low code layer sets which checks apply to whom and in what order. Decision time routing to a reviewer, a manual adjudication surface and an alternative evidence path are not described.

Independent laboratory testing under ISO 30107 3 at iBeta Levels 1 and 2 stands behind the liveness models, and the homepage states 99 percent accuracy in real time verification. That figure has no method or test set, and face comparison error rates and anything on the credit scoring models are unpublished.

Allo Bank onboarded more than 750,000 customers during its launch, a volume figure rather than an outcome against a baseline, and Nanovest describes secure onboarding in sixty seconds.

Alternative scoring improves with pooled repayment outcomes and verification with accumulated captures, and competing banks and lenders share the platform. Use of one client's applicants in models serving another, or of biometric captures in a training corpus, is never stated.

A website level privacy policy, updated September 19, 2025 and naming Google Analytics and HubSpot, does not reach the verification service. Where a facial template rests and how long it survives a completed verification are not stated, and no retention schedule or deletion commitment for identity documents appears.

ISO 27001, a privacy information management certification and a BSI certificate are shown as images without numbers, dates or scope. Encryption, key management and separation of biometric captures between institutions are left undescribed.

Journeys are built around each market's local know your customer, know your business and money laundering requirements, without naming any specific regime or statute. Its homepage claim of 100 percent compliance is an assertion, not a register entry.

Faces are verified across Southeast Asia, South Asia and China and applicants scored on alternative data, a concrete demographic and fair lending surface. Presentation attack testing measures spoof resistance rather than equal performance, and nothing is published on demographic error rates, disparate impact or adverse action.

Journey configuration and thresholds give the institution control and a basis to challenge results. A person whose face fails to match or whose document is judged forged has no described route to an alternative check or human review, and no guarantee or indemnity is published.

Advance Intelligence Group's acquisition of Jewel Paymentech brought merchant due diligence in house, shortening the chain. Model providers for biometric, document and scoring components and the sources behind screening and alternative data are unnamed.

Institutions connect their own data sources and assemble workflows without engineering through the orchestration platform, documented at doc.advance.ai. No core banking, loan origination or case management system is named, so how a result and its audit trail reach examiners is unclear.

Data may be transferred to and stored outside Singapore with no destinations named, and no hosting provider or residency commitment is published. Divergent localization rules for financial and biometric data across the markets served make that gap concrete.

Prospects go through a demo request, with no public trial, self service signup or published price. Whether identity, screening and merchant risk are charged per verification, per module or as a license is left open.

Banks such as Standard Chartered and CIMB Niaga, digital banks, lenders and crypto platforms make up the financial base, and merchant due diligence extends coverage to marketplaces. Reach spans Southeast Asia, India and mainland China, with retail and telecoms beside financial services.