Evals for Strise

Auto-clear vs. escalate routing

Correctly clears routine low-risk cases and escalates cases the policy marks as needing judgement, without silently auto-approving an escalation trigger.

Case file assembly for complex cases

Screening results, ownership, documents, and risk view arrive pre-assembled in one record so the analyst reviews rather than builds the file.

Entity vs. individual onboarding paths

Handles KYB business verification and KYC individual verification with the checks appropriate to each, including registry lookup for entities.

Case state and stall handling

Tracks open cases through waiting-on-client states rather than leaving them idle in a queue with no owner or follow-up.

Upfront document requirement mapping

Determines the full document set for a case in one request instead of discovering requirements serially across multiple client round-trips.

Submission-time validity checks

Detects expired, wrong-page, or unreadable submissions at upload and asks for a correction rather than passing them to review.

Automated follow-up on non-response

Tracks whether a request was opened and submitted, and triggers chase behaviour without an analyst noticing the stall.

Extraction into the entity record

Verified document data lands in the structured record traceable to the source document, without a manual copy-paste step.

Cumulative and aggregate control logic

Applies combined holdings against the UBO threshold rather than testing each shareholder individually in isolation.

Cross-border chain traversal

Follows ownership through multiple registries and jurisdictions, and states where a chain is partial rather than presenting it as complete.

Registry gap handling

Flags incomplete or missing registry data explicitly and routes to client outreach instead of inferring a structure from fragments.

Opaque structure termination

Recognises trusts and foundations as chain endpoints requiring declared controllers and surfaces the unresolved gap for analyst action.

Consistent policy application

Two equivalent entities receive the same classification, so the score reflects the case rather than the reviewer.

Event-driven recalculation

Recalculates on material change — director change, ownership shift, sanctions hit — rather than only at scheduled review.

Factor breakdown and source citation

Exposes which factors drove the score and cites the underlying source for each, sufficient to defend the decision later.

Policy change propagation

Updated thresholds or new jurisdictions take effect in scoring rather than diverging from what analysts actually apply.

PEP and sanctions hit adjudication

Cross-references date of birth, nationality, and jurisdiction rather than treating a name match as a conclusion.

False-positive suppression with memory

A previously dismissed, unchanged hit does not consume the same review time again in the next cycle.

Material change detection between reviews

Surfaces ownership, director, and screening changes as they occur instead of at the next scheduled review date.

Analyst decision feedback into the record

Judgements made on a case persist into the entity record so the following review starts from that position.

Portfolio-wide gap analysis

Identifies which files are stale or incomplete across the whole portfolio before any single case is opened.

Risk- and severity-based ranking

Orders remediation by regulatory exposure and gap severity rather than by visibility or ease of completion.

Bulk outreach without relationship managers

Triggers re-KYC requests at portfolio scale so senior staff are not absorbed into administrative follow-up.

Reconstructable audit trail

Records what was checked, when, and by whom, so proving completion does not require rebuilding the project from emails.