Rori Agent OS
Rori is the runtime every Roe agent runs on. It reads your data, follows your SOPs, and ships dispositions your QA, partner bank, and examiner sign off on.
The stack
20+ native connectors. Warehouse, case manager, IDV, screening, support desk.
Policies are first-class artifacts. Every disposition cites the step it satisfies.
Every fact links to the field, document, or query that produced it.
Every prompt, source, and decision logged immutably, exportable to your GRC.
Clears identity escalations: document review, selfie mismatch, data-vendor conflicts, and step-up outcomes.
Reads
Follows
Your KYC program, CIP thresholds, and step-up escalation SOP.
Ships
A verify / step-up / reject decision with the mismatch explained field by field, ready for QA sign-off.
Underwrites and monitors business customers: KYB, beneficial owners, web presence, prohibited activity.
Reads
Follows
Your merchant and business-onboarding policy, category rules, and RFI playbook.
Ships
An approve / decline / RFI disposition across ten-plus risk dimensions, cited to every source it read.
Works the fraud queue: disputes, chargebacks, ATO, and first-party versus third-party calls.
Reads
Follows
Your fraud SOP, typology definitions, and customer-communication scripts.
Ships
A cited disposition with the typology label your models train on, plus the customer message drafted.
Investigates transaction-monitoring alerts end to end and writes SAR-ready narratives.
Reads
Follows
Your AML program, alert-handling SOP, and SAR narrative standards.
Ships
A close or escalate decision with a fact-grounded narrative no examiner mistakes for a template.
Adjudicates screening hits: sanctions, PEP, and adverse-media matches, true or false positive.
Reads
Follows
Your screening policy, match-scoring rubric, and escalation thresholds.
Ships
A true / false-positive adjudication with the disambiguating evidence quoted and linked.
Shared guarantees
Because every agent runs on the same OS, adding the next queue is configuration, not a new deployment.
Dispositions with cited evidence, on every agent
Median runtime per case
Audit trail across all five agents
Atlas typology knowledge on every case