Complete work papers generated before your team logs in
Arvexi Cortex doesn't just flag accounts. It documents them. Three format types (balance comparison, account analysis, variance analysis), AI-generated narratives for accounts that need attention, standard tick marks, and living documents that grow with every balance change and event. GREEN accounts get template-based papers at zero AI cost. AMBER and RED accounts get full AI narratives with cited evidence.
The problem
Your team spends more time documenting the investigation than doing it.
The investigation is half the battle. After finding the root cause, your accountants still need to write up the findings, format the work paper, attach supporting evidence, apply tick marks, and submit for review. For 200 accounts, that's hundreds of hours of documentation work every close cycle. Much of it formulaic and repetitive.
Three format types
The right work paper structure for every reconciliation method
Balance Comparison
Compares GL balance to subledger or bank statement. Documents the source balance, comparison balance, variance amount, and explanation. Used for bank reconciliations, subledger-to-GL reconciliations, and any account with a clear external or subsidiary reference point.
Account Analysis
Breaks down the ending balance into component items. Each component is listed with its amount, description, and supporting evidence. Used for prepaid accounts, accrued liabilities, and balance sheet accounts where the balance is justified item by item rather than by comparison.
Variance Analysis
Compares the current period balance to prior periods and explains the flux. Documents period-over-period changes, identifies the transactions driving the variance, and classifies each change. Used for P&L reconciliations and accounts where the trend matters more than the absolute balance.
AI narratives
Claude Sonnet writes the analysis your team used to write manually
For AMBER and RED accounts, Cortex generates a structured narrative using Claude Sonnet. The narrative explains why the account scored the way it did, references specific transactions and balances from your database, and recommends where to focus the review. For RED accounts, Cortex additionally generates anomaly reasoning. A separate section that explains specifically why the account failed confidence checks and what makes it unusual compared to prior periods.
Tick marks
Five standard tick marks with full audit trail
Every reconciling item and work paper section receives a tick mark: AR (auto-reconciled, passed without human intervention), MC (matching coverage, verified by transaction matching), AJ (adjustment, correcting entry or reclassification), RI (reconciling item, known timing or permanent difference), CM (comments, preparer or reviewer notes). Each tick mark is timestamped and attributed to the user or system that applied it. Your auditors see a complete chain of evidence.
Living documents
Work papers grow with every event | they never go stale
Work papers are not generated once and forgotten. The appendUpdate mechanism supports 6 update types: balance refresh (new GL or subledger data), finding added (investigation discovered something new), item created (reconciling item added), status change (workflow state transition), narrative update (AI analysis refreshed), and reviewer comment. Every update is timestamped and attributed. Upsert semantics ensure Cortex can safely regenerate any section without creating duplicates. Part of Cortex. See also Work Papers in Account Reconciliation.
