When a purchase receipt is validated using Artificial Intelligence, three possible actions may occur:

1. Automatic rejection: the receipt does not meet the minimum requirements (the receipt does not include the store information, date, or total amount are not detected, or it is a duplicate).
2. Automatic validation: the receipt meets the minimum requirements and also the specific conditions defined by the administrator and is therefore valid. The user can participate in the promotion. 
3. Sent for manual review: the receipt meets the minimum requirements but does not meet one or more of the specific conditions defined by the administrator.

This guide is intended for agents responsible for manually reviewing receipts to decide whether they should be accepted or rejected.

Manual review exists because:

• Receipts uploaded by users may have poor quality or low legibility.
• AI detection is not 100% accurate, especially on long or blurry receipts.

In this guide, you will learn how to:

• Identify the exact reason why a receipt has been sent for manual review.
• Apply the correct action to approve or reject the receipt based on the detected reason.

The guide is organized as follows:

1. Identifying the receipt issue

   • Step 1: Locate the conditions that are not met
   • Step 2: Compare the information extracted by AI with the receipt photo

2. Receipt resolution and validation

Case 1. Issue with the receipt date

Case 2. Issue with store detection

1. Store name detected incorrectly

2. Correct store name, but not listed in the tag dictionary

Case 3. Issue with Product Tag

1. Purchase line not detected by AI

2. Article description detected incorrectly

3. Correct purchase line, but not associated with the article tag

1. Identifying the receipt issue

There are three main situations where a receipt may remain pending and not be automatically approved:

1. Issue with the receipt date: The date extracted by AI does not match the one on the receipt.
2. Issue with store detection. Two different situations may occur:

   • The AI has incorrectly detected the store name.

     The AI has correctly detected the store name, but it is not included in the tag dictionary.

3. Issue with product detection. Three different situations may occur:
   • A purchase line has been missed and not detected by the AI.
   • The description, quantity, or amount of the product was detected incorrectly.
   • The purchase line is correct, but not associated with the article tag.

The manual receipt review tool makes it easier to identify the issue and make decisions.

Step 1. Locate the receipt conditions that are not met

• Access the manual review tool.
• In the "Under Review" tab, click on the receipt you want to review and moderate.

When you access the pending receipt view, you will immediately see the reason why the specific receipt selected has not been approved. On the right side of the screen, you’ll see the conditions that the receipt did not meet.

For example, in the following screenshot we can see a receipt pending review because it does not meet the “Receipt Date” condition:

Step 2. Comparte the information extracted by AI with the receipt photo 

Once you’ve located the condition that the receipt did not meet, the next step is to understand why it was not met.

To do this, compare the photo of the receipt with the information extracted by the AI. By reviewing both, you’ll be able to identify the specific cause of the non-compliance.

The issue may be due to three reasons:

1. The AI omitted the information: You can locate a date in the receipt image that does not appear in the information extracted by the AI.
2. The AI modified the information: The data appears, but with errors. For example, the image shows the date 10/10/2025, but the AI detected 10/10/2023.
3. The AI correctly detected the information, but the tag is incomplete. For example, the AI detects the text “White Crop Top XL”, but the tag configured for that "T-shirt" article does not include this description.

To compare the receipt photo with the information extracted by the AI, click the "View AI output" button:

 The following view will appear:

• On the left, the photo of the receipt uploaded by the user with zoom.
• On the right, the information automatically extracted via Artificial Intelligence.

This comparative view will allow you to contrast the data and determine how to act in order to correct or validate the receipt. Pay special attention to:

• Receipt date.
• Store name.
• Purchase lines and item descriptions.

Structure of the information extracted by the AI (JSON format):

The following diagram details the structure of the information provided in JSON format:

2. Receipt resolution and validation

Once the reason why a receipt is pending has been identified, the next step is to apply the corresponding corrective action in order to approve it.

The system provides three review tools that you can use depending on the type of issue detected:

1. Improve tag rules: Use this when the AI has correctly detected the information, but the configured tag is incomplete or does not cover all possible variations of the data (for example, item descriptions or store names). 
2. Edit receipt: Allows you to adapt the information extracted by the AI. Use this when the AI has detected a data incorrectly, such as an incorrect date or a modified product reference number. It allows you to manually adjust the detected information.
3. Add article: Allows you to add information to the receipt. Use this when the AI has omitted part of the receipt content, such as undetected purchase lines or missing items. This tool allows you to manually add the missing information.

Below, we explain the steps to follow depending on each case.

Case 1: Issue with the receipt date

Problem: The date extracted by the AI does not match the date on the receipt. This can happen if the AI could not detect the date correctly, due to reasons such as the receipt being wrinkled, the photo or the printout of the receipt being of poor quality, etc.

Resolution: Edit the receipt date manually to match the one in the image. To do this, follow these steps:

• Click the "Edit receipt" button
• Modify the value of the date detected by the AI to the correct value.
• Click the "Finish editing" button.
• Click the "Reprocess" button. The system will correct the receipt information, for you to be able to validate it.

You can see it step by step in the following image:

Case 2: Issue with store detection

Below, we explain the two situations in which a receipt remains pending due to an issue with store detection:

• 1. Store name detected incorrectly

• 2. Store name detected, but not in the Receipt Store Tag

1. Store name detected incorrectly

Problem: The AI has detected the store name incorrectly. This can happen if the receipt is wrinkled, or the photo is of poor quality, and the AI couldn't detect the store name correctly.

Resolution: Edit the receipt and correct the store name detected by the AI. To do this, follow these steps:

• Click the "Edit receipt" button.
• Modify the store name detected by the AI to the correct one.
• Click the "Finish editing" button.
• Click the "Reprocess" button. The system will correct the receipt information.
• Once reprocessed, the "Accept" button will appear to validate the receipt.

In the following screenshot, we show the steps in detail:

2. Correct store name detected, but not listed in the the Store Receipt Tag 

Problem: The AI has correctly detected the name of the store, but the exact wording or formulation was not included in the Store Receipt Tag.

Resolution: Add the store name to your Store tag dictionary so that the AI can recognize it correctly. To do this, follow these steps:

1. Click the "Improve tag" button.
2. Select the tag you want to improve. Here, you should select the Store Tag that you configured in the condition.
3. Select the non-matching line or manually add the new value that should be validated in all receipts. 
4. Click the "Add value" button.
5. Click the "Reprocess" button. 
6. Finally, click the "Accept" button to validate the receipt.

In the following screenshot, we show the steps in detail:

Case 3: Problem with article detection

Below, we explain in detail the three situations where a receipt has been left pending due to a problem with article detection:

• 1. Purchase line not detected by AI

• 2. Article description incorrectly detected

• 3. Correct purchase line, but not associated with the article tag

1. Purchase line not detected by AI

Problem: The purchase line does not appear in the receipt information extracted by AI. This is a common case with photos of very long purchase receipts, where the AI is unable to extract all the purchase lines with the details of products bought.

Resolution: Manually add the purchase line to the receipt and the article tag. To do this, follow these steps:

1. Click the "Add item" button
2. Next, complete the item information (Name, Quantity, and Amount as the minimum required fields).
3. Click "Add item".
4. Click the "Reprocess" button. The system will correct the receipt information.
5. Once the receipt is reprocessed and if valid, click the green "Accept" button to validate the receipt.

In the following screenshot, we show the steps in detail:

2. Article description incorrectly detected

Problem: The purchase line appears in the receipt information extracted by AI, but the description detected does not match 100% with the information shown in the image of the purchase receipt. This can happen if the receipt is wrinkled, or if the photo quality is poor, and the AI was unable to correctly identify the product information.

Resolution: Edit the receipt and correct the products and articles detected by AI. To do this, follow these steps:

1. Click the "Edit receipt" button.
2. Modify the detected article value by replacing it with the correct value.
3. Click the "Finish editing" button.
4. Click the "Reprocess" button. The system will correct the receipt information.
5. Finally, once the receipt is reprocessed and if valid, click the green "Accept" button to validate the receipt.

In the following screenshot, we show the steps in detail:

3. Correct purchase line, but not associated with the article tag

Problem: The product line has been detected and matches 100% with the information shown in the image of the purchase receipt, but the article does not appear in the Article Tag dictionary.

Resolution: Add the article name to the tag dictionary so that the AI can recognize it correctly and validate this receipt as well as any other future receipt uploaded by clients. 

To do this, follow these steps:

1. Click the "Improve tag" button.
2. Select the tag you want to improve. Here, you should select the Article Tag that you configured in the condition.
3. Select the non-matching line or manually type in the new value to the tag rule.
4. Click the "Add value" button.
5. Click the "Reprocess" button.
6. Finally, click the "Accept" button to validate the receipt.

In the following screenshot, we show the steps in detail: