Who sees what
The guarantees underneath
Trust you can verify.
Confidentiality computed once
Every external surface — client portal, vendor portal, reports, PDFs — runs through one confidentiality filter that computes the safe subset in exactly one place. There is no seventh copy of the rule to forget to update.
One rupee, one home
Every monetary value has a single source of truth; everything else is a derived view or a linked reference with a reconciliation job that detects divergence. Money can’t quietly disagree with itself.
No guess without a loop
Every AI decision ships checkable: the human’s confirm/edit is captured, guesses are scored against outcomes, and accuracy feeds back. Nothing that moves work or money is AI-decided — a human confirms every completion, QC and payment.
Everything on the record
Every money or status change is written to an audit log. Payment release uses maker/checker separation, requires a UTR/cheque reference, and is guarded so a double-release is impossible.
Tokens that expire and revoke
External access is via a single token service — expiring, revocable links, uniformly validated. No portal link works forever, and any of them can be pulled.
A published AI benchmark
Drawing Q&A is tested against a golden set of real questions from real tender drawings, and we track that number openly — it’s a score we improve against, not a marketing claim. Every answer cites the drawing page it came from.
Under the hood: row-level security everywhere, org-scoped lookups, financial-document gating (engineers can’t open a signed work order even through drawing search), atomic money-path guards, and a per-org AI spend cap.