FAQ
Questions, answered.
How is consent handled for GST, ITR and bank data?+
Consented fetches (GST OTP fetch, Account Aggregator, ITR) are captured in-flow with the subject's authorised signatory, stored with expiry, and logged for DPDP compliance. Reports clearly mark which sections used consented vs public data.
Do LLMs decide the credit score?+
No. The score comes from a versioned, deterministic 6-pillar model. Language models only write narratives — and every claim they make must cite persisted evidence or it's rejected by the citation gate.
Which entity types are covered?+
Private limited companies, LLPs, partnership firms and proprietorships. Coverage per section adapts to the entity type — e.g. MCA data for companies/LLPs, GST + bank data for proprietorships.
How long does a report take?+
Typically under 15 minutes for all nine sections — agents run in parallel, and the longest pole is usually the consented data fetch. You can watch each section stream in live.
Can we integrate it with our LOS or ERP?+
Yes — a REST API (create/list/fetch reports, idempotency keys) plus HMAC-signed webhooks for report.completed, report.finalized, monitoring.alert and credits.low.
What can analysts override?+
Any pillar score, before finalizing — with a mandatory written reason. Overrides are visible on the scorecard and preserved forever in the audit trail. The AI's original score is never erased.
How does monitoring work after the report?+
Watchlisted subjects are swept daily or weekly (0.1 credit per subject per sweep) for GST filing lapses, NCLT petitions and rating actions. Alerts appear in the inbox and fire monitoring.alert webhooks.
Where does our data live?+
In India, on AWS (Mumbai region) — DocumentDB for structured data, S3 (SSE-KMS) for documents, Bedrock for AI. Org-scoped isolation, immutable audit trails, no training on your data.