> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.monite.com/v-2024-05-25/expense-management/receipts/ai-receipt-automation/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.monite.com/_mcp/server.

# AI receipt automation

## Overview

AI receipt automation eliminates manual expense processing by automatically matching uploaded receipts to existing transactions, assigning cost centers, tax rates, and general ledger accounts, and generating expense descriptions.

When receipts are uploaded or OCR processing completes, the system analyzes receipt content, merchant information, transaction data, and user context to populate expense fields without manual intervention.

The system performs three automated tasks when receipts are processed:

* [AI receipt-to-transaction matching](#receipt-to-transaction)
* [AI auto-classification](#auto-classification)
* [AI-generated expense descriptions](#expense-description)
* [AI auto-tagging (Coming soon)](#auto-tagging)

### AI Receipt-to-transaction matching \[#receipt-to-transaction]

Receipts can be automatically matched to transactions by an **AI-driven matching engine** once the OCR is completed. The system evaluates only **unmatched items** with the same `entity_id`, a timestamp difference of ≤1 day, and either the same `(amount + currency)` or `(merchant_amount + currency)` (with ≤1% allowed difference).

The AI logic applies a two-step approach:

1. **Equal match** - Exact merchant name match (case-insensitive, trimmed).
2. **Semantic match** - If no exact match, merchant name + location are embedded and compared using fuzzy semantic similarity (≥0.8).

#### Considerations

* The matching process runs only when receipts and transactions share the same `entity_id` and satisfy the required timestamp and amount conditions.
* When an exact merchant name match is found, the receipt is immediately linked to the transaction.
* If there is no exact match but the semantic similarity between merchant name and location is at least 0.8, the system also links the receipt automatically.
* If neither condition is met, the receipt remains unmatched.
* To ensure data consistency, each receipt can only ever be linked to one transaction.

### AI auto-classification \[#auto-classification]

When a receipt is uploaded or OCR processing completes, Monite's AI automatically populates the `cost_center_id`, `tax_rate_id`, and `general_ledger_id` fields without requiring manual input.

The AI classification uses intelligent matching to assign the correct accounting codes:

**Tax codes**

* Matches against `name` and `description` fields.
* Validates selections using `total_tax_rate` for arithmetic accuracy.

**General ledger accounts**

* Matches against `name` and `description` fields.

**Cost centers**

* Matches against `name` and `description` fields.

The AI analyzes multiple data points from the receipt to determine the best match:

* **Receipt category** (primary signal): meals, travel, office supplies, etc.
* **Line item details**: item names and descriptions
* **Merchant information**: name, description, and business category
* **Transaction metadata**: date and location

### AI-generated expense descriptions \[#expense-description]

After OCR processing, receipts are automatically enriched with **AI-generated descriptions**. The system analyses both the OCR output and the transaction context to populate the `description` field with concise, business-relevant content.

The AI interprets receipt details to summarize what was purchased and, when possible, the business purpose or context. This reduces manual entry by providing pre-filled descriptions that users can review and adjust as needed.

For example, a restaurant receipt from *The Blue Door* in London at 6pm could be described as **"Dinner in London"**, while a taxi receipt from the airport to the city center during a business trip might become **"Taxi ride"**. If the AI cannot confidently generate a meaningful description, the field remains empty instead of filling in potentially misleading text.

### AI auto-tagging (Coming soon) \[#auto-tagging]

On receipt upload or OCR completion, AI will automatically suggest the Cost Center, Tax Rate, and General Ledger account. Tags are based on receipt data and context, reducing manual work. If no confident match is found, fields stay empty for user input.

> Use AI to automatically match receipts to transactions and assign expense categories.