Agentforce Activation Workshop Exercise Guide
Agentforce Activation Workshop Exercise Guide
Exercise Guide
Exercise Guide
Agentforce Activation Workshop
Instructions:
a. From the Setup menu, select Setup. The Setup page opens in a new tab.
b. In the Setup Quick Find, search for and select Einstein Setup.
c. Select the Turn on Einstein toggle, and make sure Einstein is On.
b. Select the Einstein Copilot for Salesforce toggle, and ensure it's set to On.
b. Select Get Started. This initiates the automated setup process. Depending on the
amount of data in your org, this may take 20 minutes or more to complete. It’s
complete when you can see your Tenant Endpoint displayed.
c. Notice that you can assign user permissions here to control who can access specific
Data Cloud features and data.
In this exercise, you’re going to create an Agentforce Service agent that can assist your customers
with support queries and escalating issues. Here are the tasks you’re going to complete:
1. Create a new Service Agent to deal with support queries and escalation.
2. Review the actions and guardrails associated with the Case Management and Escalation
topics.
Instructions:
Task 1: Create a new Service Agent to deal with support queries and escalation.
2. In the Setup Quick Find, search for and select Agents. Notice there is an out-of-the-box
agent, Einstein Copilot, already set up for you.
Note: If the New Agent button isn't present, refresh the page until the button appears. If it
doesn’t appear after the refresh, you may need to request an Agentforce Service Agent
permission set license for your org. In the meantime, instead of creating a new agent, you
can amend the Einstein Copilot agent.
5. Select Next.
6. Unselect the topics listed by selecting the Added button for each, so that Case
Management and Escalation are the two remaining topics.
7. Select Next.
c. Company: [Add a brief description of your company. Use the example below for
guidance.]
Coral Cloud Resorts provides customers with exceptional
destination activities, unforgettable experiences, and
reservation services, all backed by a commitment to top-notch
customer service.
e. For now, leave Enrich event logs with conversation data unchecked.
f. Select Next.
g. Select Create.
Note: Agent users require the Agentforce Service Agent User, Agentforce Service
Agent Base Access, and Agentforce Service Agent Object Access
permission sets in order to function autonomously. To assign the required
permissions to the new agent user, edit the agent user like any other user.
Task 2: Review the actions and guardrails associated with the Escalation and General FAQ
topics.
1. With your new My Service Agent open in Agent Builder, go to the Topics workspace by
selecting the # in the vertical toolbar. Notice there are two topics assigned to your agent:
Case Management and Escalation.
2. Select the Escalation topic. Review the Classification Description, Scope, and Instruction.
These are used by the reasoning engine to determine where to direct the customer’s
request for processing. Notice the details are read-only. This is an out-of-the-box topic and
to make any amendments you must create a new version.
4. Select the Case Management topic. Notice this topic also has the Classification
Description and Scope specified. However, it has more instructions, guiding the agent in
how to complete the customer’s request.
5. Select the This Topic’s Actions subtab. Notice the action or actions required to complete
responses are listed here. Check you have only these actions: Get Record Details, Identify
Record by Name, and Summarize Record. If you don’t have these, the Agent won’t be able
to retrieve or summarize your cases. To add them:
a. Return to the Topic Configuration tab and select New Version to create a new
version of the topic.
d. Select the three missing actions from the Asset Library then Finish.
e. To remove any additional Case Management actions you don’t need, select the
menu beside the action and choose Remove from Topic.
g. Select Save.
1. With your new My Service Agent open in Agent Builder, select the Refresh icon in the
Conversation Preview pane.
2. Return to the Case tab and make a note of the Contact Name and Contact Email for the
case record you opened earlier.
3. Switch back to the Agent Builder tab. In the prompt box at the bottom of the Conversation
Preview type this question:
My name is [add an existing contact’s name]. Can you show me all the
cases I currently have open?
4. In the central pane, you can see how the agent’s reasoning engine reviews the customer
query (user prompt), selects the appropriate topic, and works out which instructions and
actions to follow.
5. The agent decides that the contact isn’t already known to them, so it asks them to provide
their email address. Add the email address when requested, and then press Enter to
continue.
6. The agent provides a summary of the contact’s current open cases and then asks if there’s
anything else they want to do with the cases.
7. In the prompt add: For my [reference the subject details of one of the cases here] case
please add a note that we are looking for some compensation.
8. In the central pane, notice the agent has selected the Case Management topic, and then
reviews the associated instructions and actions to help it decide what to do. It then
successfully adds a comment to the correct case.
c. Select the All Updates subtab. Notice there was a new comment added a few
minutes ago: “We are looking for some compensation.”
11. In the Conversation Preview prompt add: I would like to escalate my upgrade
request.
12. The agent switches to the Escalation topic and informs you that it is connecting you to a
service representative.
1. Once you’ve confirmed that the agent is performing as expected, select Activate above the
Conversation Preview to activate the agent.
2. Select ← in the Agent Builder header to exit Agent Builder and return to Setup.
Exercise 2: Extend your agent to manage customer reviews with a custom topic and actions.
In this exercise you’ll create a topic and associated actions to manage customer reviews. If you
don’t already have a customer review object in your org, you can create one now.
5. Add the new custom action to the Review Management topic and activate the agent.
Instructions
(Complete this only if you don’t have an appropriate object to use in your org.)
f. Select Launch New Custom Tab Wizard after saving this custom object.
g. Select Save.
4. Select the Fields & Relationships workspace and add fields to the object.
i. Select New.
i. Select New.
iv. Length: 1
vii. Select Always require a value in this field in order to save a record and then
select Next.
i. Select New.
d. Go to one of your contact records and add two reviews in the Customer Reviews
section of the Related tab.
In this task you create a flow that can be used by your service agent to create customer reviews.
e. Open the Resources panel by selecting the sidebar icon to the left of the Select
Elements button.
f. Using the New Resource button, create five variables. Each should have a Resource
Type of Variable. The remaining configuration details are shown below.
g. Mouse over the circle after the Start element, select +, and select Create Records.
vii. Check Manually assign variables; this will allow you to save the Id of the
created record in a chosen variable.
j. Select the back arrow in the Flow Builder header to exit the builder and return to
Setup.
5. Select Next.
6. Leave the default instructions in.
7. For each of the three inputs you want to collect from the user (comment, rating, and
reviewName) check Require Input and Collect data from user.
8. For the input contactId, select only Require input.
9. Check Show in conversation for the Id output.
Actions are grouped under topics to provide additional guidance and guardrails for the LLM. In this
task, you’re going to create a new Customer Reviews topic with the appropriate classification,
scope, and instructions before adding the action you’ve just created to the topic.
5. From the toolbar on the left, select # to open the Topics workspace.
8. Select Next.
Task 5: Add the new custom action to the Review Management topic and activate the agent.
1. From the list of Agent Actions, select the checkbox next to the Create a Review action you
just created.
2. Select the checkbox next to the Identify Record by Name action. If you don’t have this
action in your org, select Identify Contact by Email.
3. Select Finish.
2. In the prompt box introduce yourself by name using a contact you already have on the
system, then add: I would like to provide some feedback.
3. The agent thanks you personally for helping improve the business, then requests the review
comment and rating.
4. In the prompt box add: I'd like to give a rating of 5 stars for the
amazing service I received from the team last week when rearranging
my delivery.
5. If you get asked to provide your contact Id, tell the agent you don’t have it and ask if they
can find it. The agent will go off and find it, then create your review.
In this example, the service team at our company has, over time, developed a document that
outlines responses to the most commonly asked questions they receive from customers. You will
add the document to the data library for your My Service Agent and create a topic, prompt
template, and action to use the information.
If you have a user guide or similar document in PDF format, you can use that instead to inform
your agent.
3. Write a prompt template using the data from the FAQs PDF.
5. Create a custom action called Answer Questions using FAQs that use the prompt template
you created earlier.
Instructions
2. In Setup, use Quick Find to locate and open Einstein Data Library.
b. Name: CC FAQs
e. Select Save.
a. Drag and drop Coral Cloud Resorts FAQs.pdf (or your selected PDF) from your
desktop to the file upload. Or, select Upload Files, navigate to the PDF, and select it.
6. Use the App Launcher to locate and select the Data Cloud app.
a. In the Data Cloud app, select the Search Index tab. (You may need to select it from
the More dropdown menu.)
e. Change the Search Index Configuration Name to FAQUDMO and then select Save.
Once the Search Index process is complete, you will be able to view the data in Data Explorer.
3. Once the data is indexed you will also be able to view the data in Query Editor.
a. In the Data Cloud app, select the Query Editor tab (this may be under the More
menu).
b. Select New.
Note: If your query result is empty, don’t worry! It can take up to 30 minutes for indexing to
complete. Come back later and try again.
Task 3: Write a prompt template using the data from the FAQ PDF.
4. Select Next.
QUESTION_QUERY
Please use the following information to help you answer the question:
FAQ_RECOMMENDED_ANSWERS
6. Replace the placeholders with the corresponding merge fields (select the Resource search
box to access merge fields):
c. Number of Results: 6
b. Enter a relevant question into the Question field, such as: What time is
check-in?
9. In the Resolution panel, examine the prompt that was generated. The JSON structure
represents the weighted responses from Einstein Search.
10. In the Response panel, examine the response the LLM generated.
11. Enter another prompt in the Input field, such as: Is there a hairdryer in my room?
12. Select Preview and then select Activate. This prompt template can now be used in a
custom action in your agent.
h. In the Instruction box add: If you are unable to help the customer even
after asking clarifying questions, ask if they want to escalate
this issue to a live agent.
8. Select Next.
Task 5: Create a custom action called Answer Questions using FAQs that use the prompt
template you created earlier.
5. Select Next.
9. Select Finish.
The agent is ready to answer questions based on your FAQ information. Test it before activating it:
1. In the Conversation Preview prompt, type a relevant question, such as: Are there
hairdryers in the room? Then press Enter.
2. Now check that the other topics still work. Type a question relevant for your previous topic,
such as: Do you have any snorkeling experiences available? Then press
Enter.
3. Spend some time trying other prompts. Once you’re confident the agent is performing as it
should, select Activate.