Description Origin and Human-in-the-Loop
When AI writes a description, the system knows AI wrote it — and does not publish it until a human signs off.
Framework / standard: ISO/IEC 38507 §6.3 · ISO/IEC 42001 A.10.4
A human-in-the-loop without a sign-off is not human-in-the-loop. Linedat marks the origin of every asset, column or term description, and records who approved or rejected it, with what role and why.
Explicit origin, never confused
Every description carries its origin: human, AI draft, AI-reviewed or approved. An AI-generated description is marked as a draft and is never confused with human content. You clear the documentation backlog in hours with auto-descriptions, without signing off on what the machine generated as if a human wrote it.
Approval with an append-only trail
Every human decision (approve, reject, review) creates an independent append-only entry with the reviewer, their role at that moment, the reason and a snapshot of the content. Marking something "approved" is not enough: there is a record of who and why.
The limits (what we do not claim)
The review trail is append-only by design, but it is not protected by a database trigger like the audit chain (a direct delete at the DB level is not blocked). Single-table auto-doc does send real data samples to the model, passed through the PII sanitiser first; only bulk (batch) auto-doc is limited to column names and types.
How Linedat helps
Linedat lets you leverage AI to document without giving up control: you always know what the machine wrote and what a person signed.
Related capabilities
We do not store what you told the AI. We store that you can prove you said it.
AI Policy and Limits with Committee ApprovalWho turns on AI with confidential data is not the operator: it is the committee.
Governed Semantic SearchFind your "churn tables" even if none of them are called that — and with an audit trail.