SharinPix AI Extractor Integration in Salesforce Flow (Admin-Oriented)

Overview

The SharinPix AI Extractor allows users to analyze and extract data from images using generative AI capabilities.

The automation eliminates the need for manual data entry by extracting relevant details from uploaded images, such as receipts and proof of address, and automatically populating corresponding Salesforce fields.

This document covers the following topics:

• Use Case Overview: An explanation of how the SharinPix AI Extractor is applied to a common customer support scenario, demonstrating its efficiency in automating case record updates within Salesforce Service Cloud.

• AIExtractorAutomation Parameters: A breakdown of the input and output parameters that power the AI extraction process, detailing how the data is processed and returned.

• Implementation Guide: A step-by-step guide on integrating the SharinPix AI Extractor within a Salesforce Flow, including how to configure elements and actions for seamless automation.

• Demo Overview: AI-Driven Case Record Updates with AIExtractor Automation: A demonstration showing how a customer’s case is automatically updated with extracted information from an uploaded receipt, enhancing the case resolution process.

Prerequisites:

Before integrating the AI Extractor in a Salesforce Flow,

• Ensure that you have enabled SharinPix AI feature in your org. Follow this documentation for more: SharinPix Generative AI Analyse/Extract Image Capabilities

• Ensure that you are on a recent SharinPix Package Version. Follow this document to upgrade the SharinPix package

Use Case Overview

A retail shop selling electronics (e.g., TVs, radios, washing machines) has an after-sales website where customers can create support cases. When submitting a case, customers upload images such as:

• Purchase receipts

• Proof of address

• Photos depicting the issue with the product

Salesforce agents handle these cases within Salesforce Service Cloud. Instead of manually reviewing and inputting information from the uploaded documents, an automated process powered by SharinPix extracts key details and updates the necessary fields.

In our specific use case, the goal is to automatically extract contact details and product information from purchase receipts to enhance the case record.

AIExtractorAutomation Parameters

Input Parameters

Below are the inputs required when using the AIExtractorAutomation invocable method in a Salesforce Flow.

Parameter	Description
List of Image URLs

• This is a list containing the URLs of images to be processed by the AIExtractorAutomation.

• Each entry in the list represents the URL of an image that will be analyzed by the model.

Prompt |

• The Prompt parameter is a string that contains the question or instruction to be passed to the GPT model.

• This prompt guides the model on what information to extract from the provided images.

• It could be a general question like "What is the content of this image?" or a more specific instruction such as "Identify any text in the image."

Output Parameters

Below are the outputs received when using the AIExtractorAutomation invocable method.

Parameter	Description
List of Values (JSON list)

• This is a collection of image extraction results.

• Each result in the list represents a specific image processed by the AIExtractorAutomation class.

• Each entry provides detailed information extracted from individual images.

Raw Result |

• The Raw Result parameter holds the full JSON output.

• This is the raw data returned by the AIExtractor class, including all image extraction results.

• It contains the details for all the images sent for extraction in a single JSON response.

Implementation Guide

This section will walk you through the steps needed to create a Salesforce Flow that invokes the AIExtractorAutomation Apex class. This flow will extract contact and product details from purchase receipts uploaded by customers. It will trigger automatically when a new case is created and populate the case record with relevant details from the receipt.

Step 1: Create a New Record-Triggered Flow

To do so, follow the steps below:

• Go to Setup. In the Quick Find Box, type Flows.

• Under Process Automation , select Flows.

• Click on the New Flow button.

• Select the option Record-Triggered Flow , and click on the **Create **button.

• After clicking on Create , the _Configure Start _modal will be displayed. Fill in the modal as indicated below:

  • Select Object : Case

  • Configure Trigger : A record is created

  • For the **Set Entry Conditions **section:

    • Select None

  • When to Run the Flow for Updated Records : Every time a record is updated and meets the condition requirements

  • Optimize the Flow for : Actions and Related Records

  • Enable Include a Run Asynchronously path to access an external system after the original transaction for the triggering record is successfully committed

Step 2: Add a Get Element to get the SharinPix Image (Receipt)

Next, add a Get Element to retrieve the receipt image associated with the Case record.

For the Get Records element:

• Label: Get SharinPix Image

• Object: SharinPix Image

• Conditions:

  • sharinpix__AlbumID__c (Field) equals (operator) **{!$Record.Id} **(newly-created resource named Id)

  • sharinpix__DisplayTags__c (Field) equals (operator) **Purchase Receipt **(tag added to the Receipt Image)

• In the How Many Records to Store , check the Only the first record option

• In the Where to Store Field Values , check the** Automatically store all fields** option

• Click Done

Step 3: Add an Assignment Element to assign the receipt image URL

Create a new variable and assign the record's Image URL Full to it.

1. Go on the Manager tab and click on New Resource.

• Resource Type: Variable

• API Name: ImageUrls

• Data Type: Text

• Check the Allow multiple values (collection)

2. Add Assignment Element.

• Label: Assign Image Urls

• Conditions: {!ImageUrls} (Variable) Add (operator) **{!Get_SharinPix_Image.sharinpix__ImageURLFull__c} **(value)

Step 4: Add the Action to extract contact and product details from the receipt

Create a new variable for the prompt to be passed to the AI to retrieve the contact and product details from the receipt in a JSON format

• Go on the Manager tab and click on New Resource.

• Resource Type: Variable

• API Name: Prompt

• Data Type: Text

• Default Value: I have an automation which takes list of object with the following: client_name, product_name, payment_date, receipt_number and address_galaxy_store. I would like to extract these information provided from the images and you should return a JSON corresponding to what I can input to my automation. All values should be converted to String inside the JSON. Output only the JSON in a single-line without any markdown associated and remove any round brackets.

Create two other variables for the output obtained from the AIExtractorAutomation

1. ResultsList

   • Go on the Manager tab and click on New Resource.

   • Resource Type: Variable

   • API Name: ResultsList

   • Data Type: Text

   • Check the Allow multiple values (collection)

2. Results

   • Go on the Manager tab and click on New Resource.

   • Resource Type: Variable

   • API Name: Results

   • Data Type: Text

Add an Action element

• Search for Sharinpix__AIExtractorAutomation

• On the Action _ _ modal for Sharinpix__AIExtractorAutomation, populate the fields as indicated below:

Field	Example Value
List of Image URLs	{!ImageUrls}
Prompt	{!Prompt}

Step 5: Parse the Extracted Result

Once the AI extractor processes the receipt, the next step is to parse the extracted data (contact and product details) into separate components that can be used to update different fields in the Case record. To achieve this, we will use the FlowJSONParser invocable method as explain below:

Add an Action element

• Search for SharinPix Flow JSON Parser

• Label: Parse Results

• Set Input Values for the Selected Action: populate the fields as indicated below:

Field	Example Value
Ignore Error?	{!$GlobalConstant.True}
JSON String to Parse	{!Results}
Field01	client_name
Field02	product_name
Field03	payment_date
Field04	receipt_number
Field05	address_galaxy_store

• On the Show Advanced Options

  • Enable Manually assign variables

  • Store Output Values: populate the fields as indicated below:

Field	Example Value
Value01	{!$Record.Client_Name__c}
Value02	{!$Record.Product_Name__c}
Value03	{!$Record.Payment_Date__c}
Value04	{!$Record.Receipt_Number__c}
Value05	{!$Record.Address_Galaxy_Store__c}

Step 6: Add an Update Element to update the Case record

This step is used to update the corresponding fields on the Case record.

Add an Update element

• Select Use the IDs and all field values from a record or record collection

• Do not add any filter conditions.

• In the Select Record(s) to Update section, populate with {!$Record}(Triggering Case)

Demo Overview: AI-Driven Case Record Updates with AIExtractor Automation

In this demo, we will show how seamless and efficient the automation process is, saving time for both customers and Salesforce agents by reducing manual data entry and ensuring that all relevant case details are available at a glance.

1. Case Record Creation:

• The process begins when a customer submits a support case for the issue of their TV not turning on on the after-sales website. As part of their submission, the customer uploads photos depicting the issue with the product, the purchase receipt , which is tagged with the label 'Purchase Receipt' to easily identify it as shown in the diagram below.

2. Flow Triggered:

• Once the case record is created, the Flow is automatically triggered in Salesforce. The Flow calls the AIExtractorAutomation , which processes the uploaded purchase receipt to extract relevant details, such as contact information (e.g., client name) and product information (e.g., product name, purchase date).

3. Case Record Update:

• After the extraction process is completed, the case record is updated with the extracted information. This includes auto-populating the case fields with the extracted contact and product information, allowing the Salesforce agents to quickly access the relevant data and take the necessary next steps in resolving the case.