Configure Agentforce Service - Part2

Agent Builder:

It allows administrators to design and configure automated voice and chat agents (virtual agents) using a low-code or no-code interface. It leverages Salesforce Einstein AI capabilities to provide personalized and context-aware customer service experiences.

Now lets move to agent builder:

●?????? click on the setup

●?????? Search for agent

●?????? Click on agent

●?????? Click on create new agent

●?????? Select Agentforce service Agent

●?????? Click Next

●?????? Remove by default selected Topics.

●?????? Click Next

●?????? Give the Name and other details

●?????? Select the Einstein service agent User as the agent user.

●?????? Click create

Topics Creation(General FAQ)

●?????? Go to topics tab

●?????? create General FAQ

●?????? Select no action

●?????? Now on created topic switch to topic action

●?????? Create New action

●?????? Select action type as Prompt template

●?????? The earlier created Ecommerce FAQ will show up select it.

●?????? Provide the details : Agent action instruction : Please use this action for general query from customer. Analyse the user query using unstructured data then give relevant information

●?????? Provide the details : Agent action instruction : Please use this action for general query from customer. Analyse the user query using unstructured data then give relevant information.

●?????? Input and output as per the flow will show up.Please give the below details for input and output :

○?????? Input:EcomFAQ : This is the query asked by a customer. The query is related to some company info.Require input should be selected

○?????? Nothing change needs to be done in citation mode.

○?????? Output : promptResponse : The prompt response generated by the action based on the specified prompt and input.Select show in conversation.

○?????? Rest should be left as is.

Lets create next topic as Order Enquiries:

●?????? Don't select any action and click save.

●?????? Go to order enquiries and Topics section

●?????? Create new Topic Action consuming flows

●?????? Select action as flow

●?????? In reference action select the flow name : Get Contact by Email

●?????? Give the agent action label as : Identify Customer by email custom

●?????? Agent action instructions : Identify a customer by their email address and return their contact record.

●?????? Below are the input and output field values :

○?????? emailAddress : Email address provided by the customer.

○?????? Click require input

○?????? Output :

○?????? contactRecord

○?????? Instruction : The contact record associated with the identified customer.

○?????? Select show in conversation

Create new action for other flow :

Follow the above steps

●?????? And then? select the required flow : Fetch Order by contact Id and order Id

●?????? Give the instruction : Agent action instructions : This is the action used only when asked about order status.This action is used to fetch the order details based on the order id.

●?????? Give below inputs and outputs

●?????? contactId ?: contact is used to fetch order details

●?????? OrderDetails : this is the order id used to fetch the order details

●?????? For both select the required input

●?????? Output : Order :Order details contain info about order status. Show in conversation and object.

●?????? Finish

?

Test Scenario: Select a question from the FAQ section. The agent should provide an appropriate response.

●????? Personal Information Request: If the user inquires about personal information, such as order status, the agent must first verify the user's identity.

●????? Verification Process: Use one of the user's contact details and an associated order for verification.

●????? Order Status Retrieval: After successful verification, the agent should request the order ID.

●????? Providing the correct order ID should allow the agent to display the order details.

●????? Testing Variations:

●????? Test that a non-verified user cannot access order details.

●????? Providing an incorrect order ID should not return any order information.

●????? Additional Contextual Questions: Test the agent's ability to respond to other related queries accurately.

?