DEMO / SANDBOX

Governed autonomous intelligence

Policies, Agentic AI & RAG

Configure what the system may do, ground every consequential response and keep LLMs outside financial authority.
The LLM reasons; typed policy tools control every side effectModels cannot write balances, payments, promises or audit evidence. Every consequential response is tenant-grounded and traceable.
FAIL CLOSED
Agent runs18,49299.6% completed
Policy checks48,6127 actions denied
Grounded responses99.2%Approved sources
Knowledge sources221Tenant scoped
Tool failures4Retried safely
Unauthorized actions0All blocked

Effective policy set

Recovery Policy v6.1

Approved
Contact eligibility

Consent, ownership, verification and suppression evaluated before every contact

12 controls
Contact windows

08:00–19:30 EAT; quiet hours and tenant frequency limit enforced

8 controls
Customer protection

Hardship, dispute, wrong-number, opt-out and vulnerability fail closed

14 controls
Payment integrity

Exact amount and MSISDN consent; callback, posting and reconciliation distinct

17 controls
AI tool access

Allowlisted typed tools, strict schemas, tenant permission and idempotency checks

21 controls
WhatsApp restriction

Debt-recovery messaging cannot transition to LIVE; demo adapter only

Hard block