AI-Assisted Accounting Policy
Controls for account-mapping suggestions and payroll journal automation between ZentroPayroll and ZentroBooks.
Purpose and boundary
Automation may suggest a chart-of-accounts mapping, explain why it fits a payroll category, and prepare a balanced aggregate journal. It does not make payroll decisions, file taxes, move payroll funds, create hidden worker-level books entries, or provide professional accounting advice.
Human approval
Every suggested mapping begins in a suggested state. An authorized owner, administrator, or accounting user must inspect the account identifier, name, category, confidence, and reasoning and explicitly approve it. Sync remains disabled until every required mapping is approved. Low confidence, conflicting account type, missing totals, or an unbalanced journal blocks posting.
Data minimization
Suggestions should use aggregate payroll categories, account metadata, industry/company context chosen by Customer, and prior approved corrections. Worker bank details, tax identifiers, identity documents, and unnecessary personal data must not be sent to an AI service. Any future external model must be separately configured, documented as a subprocessor where required, and restricted from training on Customer Data unless Customer expressly agrees.
Traceability and correction
The system records suggestion source, confidence, reasoning, approver, time, mapping version, payroll source, payload, attempts, provider response, and idempotency key. A posted error is corrected through an auditable reversal or adjusting entry consistent with the connected ledger; source payroll history and prior audit events are retained.
Customer responsibility
Customer is responsible for its chart of accounts, tax treatment, dimensions/classes, cash versus accrual policy, reconciliation, financial statements, and final entries. Automated suggestions may be incomplete or wrong. Consult a qualified accountant for material or uncertain treatment.