AI Expense Coding

How auto-categorization works

When a SKOOR card transaction clears, the AI immediately assigns a category and GL code. The first signal is the Merchant Category Code (MCC) transmitted by the payment network. The AI maps the MCC to your chart of accounts, then refines the assignment using the merchant name, transaction amount, and historical patterns from your account. The result is a categorized expense before you even open the dashboard.

Learning from corrections

When you recategorize a transaction, the AI records the correction and adjusts its model. If you move a Home Depot purchase from "Office Supplies" to "Maintenance Supplies" three times, the AI learns that Home Depot transactions on your account belong in Maintenance Supplies. Corrections compound: the more you correct, the fewer corrections you need in the future. Most accounts reach 95% accuracy within 60 days of active use.

Confidence scoring

Every AI-assigned category includes a confidence score from 0 to 100. High-confidence assignments (above 90) are routed directly to approval. Low-confidence assignments (below 70) are flagged for manual review. Medium-confidence transactions are surfaced in the Smart Review queue with the AI suggestion and two alternative options. You decide the confidence threshold that triggers manual review from Settings.

Custom GL mapping

Connect your chart of accounts from QuickBooks or upload a custom GL map from Settings. The AI maps MCC codes and merchant names to your specific GL codes, not generic categories. When your chart of accounts changes, update the mapping and the AI recategorizes future transactions automatically. Historical transactions retain their original coding unless you run a bulk recategorize.