# Categorization

### AI layer

When transactions sync into Kick, the categorization model automatically categorizes \~97% of them on first pass. The model is optimized to get you to a materially correct Profit & Loss as fast as possible, minimizing manual work across all your clients.

Kick also learns from the changes you make. When you update a category, counterparty, class, or other transaction attribute on the Transactions tab, Kick prompts you to apply the same change to similar transactions — both historical and going forward. If confirmed, Kick applies that feedback automatically in the future.

Uncategorized transactions appear when the model's confidence is low and need to be reviewed manually.

***

### Rules layer

Rules give you hard, deterministic control over categorization. You can create multi-condition, multi-action rules to set categories, GL accounts, counterparties, classes, entities, and more. Rules always take priority over AI-based categorization.

Rules are especially useful for complex clients or industries that require highly detailed and consistent financial reporting — or any situation where you need to guarantee a specific outcome regardless of what the AI suggests.

→ [Rules](https://accountants.kick.co/bookkeeping-workflows/rules)

***

### How the two layers work together

For most clients, the AI layer handles the bulk of categorization automatically. Rules layer on top to enforce any exceptions, custom mappings, or client-specific logic. Transactions post to the General Ledger in **real time** once categorized.