Orula is Datab's local institutional intelligence agent. It helps staff search, summarize, draft, report, detect risk, and prepare actions — with human approval.

What Orula does

• Reads approved local institutional records to summarize operations and surface evidence.
• Detects workflow, risk, and data-quality issues using module-specific knowledge packs.
• Drafts WhatsApp reminders. The admin sends; Orula never does.
• Prepares approval briefs for sensitive actions.

What Orula does not do

• It is not an autonomous operator. Every impactful action requires explicit human approval.
• It does not approve, reject, or modify loans. It does not export sensitive data without approval.
• It does not bypass Atlas, Gu, or the audit log.

Bounded design

Orula runs behind a prompt guard, a policy engine, and a reward score that explicitly penalises unsafe behaviour. Audit logs are written for every request. The intelligence is eager; the agency is bounded.