Boundaries
What ClosePilot automates, and what always stays human-owned.
How we're different
- We gate sensitive actions with explicit approvals.
- Deterministic checks run on rails; AI drafts are optional.
- Every output links to evidence and an audit trail.
Trust rule: AI drafts, humans decide for sensitive outcomes.
What ClosePilot automates
- Evidence requests (drafts + follow-ups) under policy gates
- Reconciliation preparation and confidence bands
- Variance note drafts from structured inputs
- Audit pack assembly and export (approval-gated)
What ClosePilot does not do
- It does not post journal entries without human approval
- It does not override your accounting system of record
- It does not “auto-close” the books without sign-off
- It does not claim perfect matching or perfect classification
How we keep it safe
- Policy Studio controls what can run automatically.
- Approvals are mandatory for sensitive steps.
- If a step is uncertain, it routes to the Exception inbox.
Not legal advice. Requirements vary by institution and auditor expectations.