Persona

Document reading, face matching, passive liveness and risk scoring are model work inside the building blocks. A configurable identity workflow platform drawing results from connected providers would stand without them, and that orchestration is what the marketing leads with.

Customers author flows that vary by geography, product and risk level, deciding where automation stops, and Cases gives a dedicated review surface with role based permissions. Platform agents can now review cases in place of a team, and what they may decide alone is not described.

Because the customer authors flow logic and chooses providers, much of the decision path is known by construction. Face matching, document reading and liveness carry no published documentation, error rates or validation support.

About a billion identities verified and more than 300 million fraud attempts stopped sit beside named customers such as Square, LinkedIn and OpenAI. Outcomes on the customers page are told in words rather than figures, and no number attaches to a named customer.

Verification data is not sold, brokered or used to train AI or machine learning models, a categorical answer on the security page. Model hosting terms, supplier retention and lifecycle detail are not set out.

Identity documents, facial images and ownership records flow through the platform, with configurable automated retention and field level redaction stated on the security page. Linked subprocessor lists are absent, and the position on state biometric privacy statutes is not given.

FedRAMP authorization with a marketplace listing, SOC 2 Type II and ISO 27001 are named, with copies released through an account manager. Report dates and audit periods are absent.

Marketplace seller verification, customer identification and age assurance duties are each named specifically, alongside anti money laundering obligations for financial buyers. Holding no financial license fits a software supplier.

A bias post from October 2022, updated January 2026, states regular NIST evaluation of the face recognition models and internal audits against darker skin complexions, with a worst case single demographic false match rate of 0.00086 and false non match rate of 0.0019. Results by group with a method, and anything for liveness or age estimation, are not provided.

Step up verification routes a user who would fail a check into an additional evidence path instead of a rejection, and a customer describes it recovering users who would otherwise be lost. Vendor liability is another matter: accuracy guarantees and remediation terms do not appear.

Flows call out to third party providers the customer selects, an openly described orchestration design. Those data providers, the model providers for document reading and face matching, and a subprocessor list are unpublished.

An API, SDKs and hosted flows let a team embed verification or hand off the full experience, with public documentation, a status page and marketplace apps from the Growth plan. Core banking, digital banking and case management integrations are not named.

Google Cloud hosts the platform, with a choice of United States or European Union data centers. Where connected third party providers process documents and biometrics is not said, and an on premises option is not offered.

Essential starts at $250 a month on a 12 month minimum, charged per successful verification with abandoned attempts free, while Growth and Enterprise are quoted. The per verification rate that drives cost at volume is not published.

Financial institutions, payments and cryptocurrency each get separate material, and the bank material covers parity between branch and online channels, with Hershey Federal Credit Union named. Marketplaces, digital health and government are served too, so a bank is one audience among many.