Data Cloud and Agentforce for Admins

Part 36: Data Cloud and Agentforce for Admins

Welcome to the final post of Topic 2 in our Salesforce series. Throughout this topic we have covered every major product and platform capability an admin needs to know — from Sales Cloud and Service Cloud through Experience Cloud, automation, security, reports, and beyond. In this closing installment we tackle two of the most forward-looking features on the Salesforce platform: Data Cloud and Agentforce.

Data Cloud gives you a unified data foundation by ingesting, harmonizing, and activating data from every source your organization touches. Agentforce builds on that foundation by letting you create AI-powered agents that can reason, plan, and take action inside Salesforce — all configured declaratively.

By the end of this post you will understand the architecture of Data Cloud, know how to configure it step by step, and be able to build and deploy an Agentforce agent that handles real customer service work through Omni-Channel.

What Is Data Cloud?

Data Cloud is Salesforce’s hyperscale data platform. It sits at the center of the Salesforce ecosystem and serves as a single source of truth for all customer data — regardless of where that data originated.

The Problem Data Cloud Solves

Most organizations store customer data in dozens of disconnected systems: CRM, marketing automation, e-commerce platforms, data warehouses, mobile apps, IoT devices, and third-party services. This fragmentation creates several problems:

• Incomplete customer profiles — No single system has the full picture of a customer.
• Duplicate records — The same person exists in multiple systems under different identifiers.
• Stale data — Batch integrations mean data is always hours or days behind.
• Siloed insights — Marketing cannot see service history; service cannot see purchase behavior.
• Slow activation — Getting insights to the point of action requires manual processes.

Data Cloud addresses every one of these problems by ingesting data in near real time, harmonizing it into a common model, resolving identities across systems, building unified profiles, and activating those profiles across every Salesforce cloud and external destination.

Core Concepts

Data Cloud Editions

Data Cloud functionality is available in several packaging tiers:

To check what Data Cloud entitlements your org has, go to Setup > Data Cloud Overview or Setup > Company Information and review your feature licenses.

The Architecture of Data Cloud

Understanding the architecture is essential before you start configuring anything. Data Cloud processes data through a series of layers.

Layer 1 — Ingestion

Data enters Data Cloud through data streams. Each data stream connects to a source and defines how data is pulled in.

Supported source types include:

• Salesforce CRM — Standard and custom objects from your org (Accounts, Contacts, Cases, Opportunities, custom objects, etc.)
• Marketing Cloud — Email sends, opens, clicks, journeys, data extensions
• Commerce Cloud — Orders, products, carts, customer interactions
• MuleSoft — Any API-accessible system connected through MuleSoft Anypoint
• Amazon S3 — CSV or Parquet files stored in S3 buckets
• Google Cloud Storage — Files from GCS
• Azure Blob Storage — Files from Azure
• Ingestion API — A REST API that lets any system push data into Data Cloud in real time
• Connected Apps / Partner Connectors — Pre-built connectors for platforms like Snowflake, Google Ads, Meta Ads, and others

Each data stream deposits records into a Data Lake Object (DLO). The DLO is a raw staging table — it holds the data exactly as it arrived from the source.

Layer 2 — Mapping

Once data is in a DLO, you map its fields to a Data Model Object (DMO). DMOs follow a canonical data model based on industry-standard schemas. Salesforce provides a library of standard DMOs:

You can also create custom DMOs when none of the standard ones fit your data.

Mapping is the step where you align the semantics of your source data with the common model. For example, if your e-commerce system calls the customer email field cust_email, you map it to the Email Address field on the Contact Point Email DMO.

Layer 3 — Identity Resolution

After mapping, Data Cloud runs identity resolution to determine which records across all your data sources represent the same real-world person. Identity resolution uses rulesets that you configure. A ruleset contains match rules and reconciliation rules.

• Match rules define the criteria for deciding that two records refer to the same individual (e.g., exact match on email, fuzzy match on name plus phone).
• Reconciliation rules define which source wins when conflicting data exists (e.g., “prefer the CRM record for name, prefer the e-commerce record for mailing address”).

The output of identity resolution is the Unified Individual — a single golden profile that merges data from all matched source records.

Layer 4 — Calculated Insights and Segmentation

With unified profiles in place, you can compute calculated insights — aggregated metrics such as total purchase value, average case resolution time, or number of website visits in the last 30 days. Calculated insights are defined using SQL-like syntax or the visual builder.

You can then build segments — groups of unified profiles that share a set of characteristics. Segments are dynamic; they update automatically as underlying data changes.

Layer 5 — Activation

Activations push segments or data to downstream targets:

• Marketing Cloud — Trigger a journey or populate a data extension
• Salesforce CRM — Write back to standard or custom objects
• Google Ads / Meta Ads — Sync audiences for advertising
• Tableau — Feed dashboards and visualizations
• Amazon S3 / Cloud Storage — Export for use by external systems
• Webhook / API — Trigger external processes

This five-layer pipeline — Ingestion, Mapping, Identity Resolution, Insights/Segmentation, Activation — is the core architecture that everything in Data Cloud revolves around.

Setting Up Data Cloud Step by Step

Let us walk through a complete Data Cloud setup from scratch. We will ingest CRM data, map it, configure identity resolution, build a segment, and activate it.

Step 1 — Enable Data Cloud

1. Navigate to Setup.
2. In the Quick Find box, search for Data Cloud.
3. Click Data Cloud Setup.
4. If Data Cloud is not yet enabled, click the Enable button. The provisioning process takes a few minutes.
5. Once provisioning completes, you will see the Data Cloud Setup home page with status indicators for each component.

After enabling, Salesforce creates a set of default data streams for standard CRM objects (Account, Contact, Lead, Case, Opportunity, etc.). These are called CRM Connector Bundles.

Step 2 — Review Default Data Streams

1. From the App Launcher, open the Data Cloud app.
2. Navigate to Data Streams in the left sidebar.
3. You will see a list of pre-configured CRM data streams. Each one corresponds to a Salesforce object.
4. Click on any stream (e.g., Contact) to see its configuration — the source object, the fields being ingested, and the refresh schedule.

Default CRM data streams sync incrementally. Changes to records in Salesforce flow into Data Cloud within minutes.

Step 3 — Create a New Data Stream

To bring in data from an external source, create a new data stream:

1. Click New on the Data Streams page.
2. Select the source type. For this walkthrough, select Salesforce CRM to bring in a custom object.
3. Choose the custom object you want to ingest (e.g., Subscription__c).
4. Select the fields you want to include. Include any fields that you will need for mapping, identity resolution, or segmentation.
5. Set the refresh mode:
   • Incremental — Only changed records are synced (recommended for most cases).
   • Full Refresh — The entire dataset is reloaded each time (use for small reference tables).
6. Click Next, review your configuration, and click Deploy.

The data stream will begin its initial sync. You can monitor progress on the data stream detail page.

Step 4 — Map Data Lake Objects to Data Model Objects

Once data lands in DLOs, you need to map them to DMOs:

1. Navigate to Data Model in the left sidebar.
2. Click on the DLO you want to map (e.g., Subscription__c DLO).
3. Click Map to Data Model Object.
4. Choose the target DMO. If a standard DMO fits your data, select it. If not, create a custom DMO:
   • Click New Data Model Object.
   • Give it a name and API name.
   • Define the fields with appropriate data types.
5. Map each DLO field to its corresponding DMO field. Use the drag-and-drop mapper or the field-level dropdown selectors.
6. Designate a primary key — the field that uniquely identifies each record.
7. If the DMO relates to an individual (person), map the relationship field that links to the Individual DMO. This is critical for identity resolution.
8. Click Save.

Repeat for every DLO that needs mapping.

Step 5 — Configure Identity Resolution

1. Navigate to Identity Resolution in the left sidebar.
2. Click New Ruleset (or edit the default ruleset).
3. Give the ruleset a name and description.
4. Add match rules:
   • Click Add Match Rule.
   • Choose the match type: Exact or Fuzzy.
   • Select the fields to match on. Common configurations:
     • Rule 1: Exact match on Email Address
     • Rule 2: Exact match on Phone Number
     • Rule 3: Fuzzy match on First Name + Last Name + Mailing Address
   • Set the match priority. Rules with higher priority are evaluated first.
5. Add reconciliation rules:
   • For each field on the Unified Individual profile, define which source takes precedence.
   • Options include: Most Recent, Source Priority (you rank the sources), or Most Frequent.
6. Click Save and then Run to execute identity resolution.

Identity resolution runs as a batch process. Depending on data volume, it may take minutes to hours. After completion, navigate to Unified Profiles to see the results.

Step 6 — Explore Unified Profiles

1. Navigate to Unified Profiles from the Data Cloud app.
2. Search for a specific individual by name or email.
3. The unified profile shows:
   • Demographics — Name, email, phone, address (merged from all sources)
   • Related Records — All source records that were matched to this individual
   • Engagement Timeline — A chronological view of interactions across every channel
   • Calculated Insights — Any computed metrics attached to this profile
   • Segment Membership — Which segments this individual belongs to

This is the single-customer view that Data Cloud is built to deliver.

Step 7 — Build a Calculated Insight

1. Navigate to Calculated Insights in the left sidebar.
2. Click New.
3. Give the insight a name (e.g., Lifetime_Purchase_Value).
4. Define the calculation using the visual builder or SQL:

   SELECT
     Individual__c AS IndividualId,
     SUM(Total_Amount__c) AS Lifetime_Value
   FROM
     Sales_Order__dlm
   GROUP BY
     Individual__c

5. Set the refresh schedule (hourly, daily, or on-demand).
6. Click Save and Run.

The calculated insight is now available as a field on the unified profile and can be used in segmentation.

Step 8 — Create a Segment

1. Navigate to Segments in the left sidebar.
2. Click New Segment.
3. Give the segment a name (e.g., High Value Customers - Last 90 Days).
4. Define the criteria using the visual segment builder:
   • Drag Unified Individual into the canvas.
   • Add a filter: Lifetime_Purchase_Value > 5000.
   • Add a time-based filter: Last Purchase Date within last 90 days.
5. Click Calculate Count to preview how many profiles match.
6. Click Save and Publish.

Published segments update automatically as underlying data changes.

Step 9 — Activate the Segment

1. Navigate to Activations in the left sidebar.
2. Click New Activation.
3. Select the activation target:
   • Salesforce CRM — Write a custom field or create a list view.
   • Marketing Cloud — Push the segment to a data extension or journey.
   • Google Ads — Sync the audience for ad targeting.
   • Amazon S3 — Export as a file.
4. Select the segment you want to activate.
5. Choose the fields to include in the activation payload.
6. Set the refresh schedule.
7. Click Save and Activate.

Your segment is now flowing to its target destination on the schedule you defined.

Data Cloud Best Practices

Setting Up Data Cloud to Prep for Agentforce

Agentforce relies heavily on Data Cloud for grounding — giving AI agents the context they need to respond accurately. Before configuring Agentforce, ensure the following Data Cloud components are in place.

Checklist

Connecting Data Cloud to Agentforce

Agentforce uses Data Cloud through a feature called Data Cloud Grounding. When an agent needs to answer a question or take an action, it queries Data Cloud to retrieve relevant context about the customer, their history, and their current status.

To enable grounding:

1. Navigate to Setup > Einstein Setup.
2. Ensure Einstein is enabled.
3. Navigate to Setup > Agentforce.
4. In the agent configuration, under Data Sources, add Data Cloud as a grounding source.
5. Select which DMOs and calculated insights the agent should have access to.
6. Save and test.

The agent can now pull real-time unified profile data into its reasoning process.

What Is Agentforce?

Agentforce is Salesforce’s platform for building, deploying, and managing autonomous AI agents. Unlike traditional chatbots that follow rigid scripts, Agentforce agents use large language models (LLMs) combined with Salesforce data, business rules, and predefined actions to reason through complex tasks and take action.

Key Characteristics

• Autonomous reasoning — Agents break down complex requests into steps and execute them in sequence.
• Grounded in your data — Agents are connected to your CRM, Data Cloud, Knowledge, and other Salesforce data sources.
• Action-oriented — Agents do not just answer questions; they can create records, update fields, send emails, trigger flows, and more.
• Guardrailed — You define what the agent can and cannot do, what data it can access, and when it should escalate to a human.
• Multi-channel — Agents can be deployed in Service Console, Experience Cloud sites, Slack, and custom applications.

Agentforce Architecture

Agentforce has several core components:

How an Agentforce Agent Works

When a user interacts with an Agentforce agent, the following process occurs:

1. User sends a message — The message is received by the Agentforce runtime.
2. Topic classification — The runtime determines which topic the message relates to.
3. Instruction loading — The agent loads the instructions for the matched topic.
4. Data grounding — The agent queries relevant data sources (CRM, Data Cloud, Knowledge) to gather context.
5. Reasoning — The LLM reasons about the request, the available data, and the available actions.
6. Action execution — The agent calls the appropriate action (a Flow, Apex method, or Prompt Template).
7. Response generation — The agent constructs a natural-language response based on the action result.
8. Guardrail check — The response is validated against the defined guardrails before being sent.
9. Delivery — The response is delivered to the user through the channel.

This cycle repeats for every message in the conversation.

How to Use Prompt Builder

Before building an Agentforce agent, it helps to understand Prompt Builder — a tool for creating reusable prompt templates that agents (and other Einstein features) can invoke.

What Is Prompt Builder?

Prompt Builder is a declarative tool in Salesforce Setup that lets you create structured LLM prompts with dynamic merge fields. Instead of hardcoding prompts, you build templates that pull in record data, related lists, and Data Cloud insights at runtime.

Prompt Template Types

Creating a Prompt Template

1. Navigate to Setup > Prompt Builder.
2. Click New Prompt Template.
3. Select the template type (e.g., Flex).
4. Give the template a name and API name.
5. Select the related object — the object whose data will be available as merge fields (e.g., Case).
6. Write the prompt in the editor. The editor supports:
   • Free text — Your instructions to the LLM.
   • Merge fields — Click the merge field picker to insert dynamic references like {!Case.Subject}, {!Case.Contact.Name}, {!Case.Account.Name}.
   • Related list references — Pull in related records (e.g., all CaseComments on the case).
   • Data Cloud references — Include calculated insights or unified profile data.
   • Flow references — Call a Flow to retrieve data that is not directly available on the record.
7. Configure grounding:
   • Select Data Cloud to ground the prompt with unified profile data.
   • Select Knowledge to ground the prompt with relevant knowledge articles.
8. Set the response format — free text, structured JSON, or a specific field value.
9. Click Save.

Example Prompt Template

Here is an example of a Flex prompt template for generating a case response:

You are a customer support agent for {!Case.Account.Name}.

The customer's name is {!Case.Contact.Name}.
The case subject is: {!Case.Subject}
The case description is: {!Case.Description}
The case priority is: {!Case.Priority}

Recent case comments:
{!Case.CaseComments}

Using the above context, draft a professional and empathetic response
to the customer. Address their concern directly. If you need more
information, ask a specific follow-up question.

Keep the response under 200 words.
Do not make promises about timelines unless they are stated in the
case details.

Testing a Prompt Template

1. On the Prompt Template detail page, click Open in Prompt Builder.
2. In the preview panel, select a specific record (e.g., a Case record) to use as test data.
3. Click Resolve to see the prompt with all merge fields filled in.
4. Click Generate to send the resolved prompt to the LLM and see the output.
5. Review the output. If it is not meeting expectations, adjust the prompt text and regenerate.
6. Iterate until the output quality is consistently acceptable.

Prompt Builder Best Practices

• Be specific — Vague instructions produce vague responses. Tell the model exactly what you want.
• Provide context — Include relevant record data, related records, and historical information.
• Set constraints — Define word limits, formatting requirements, and topics to avoid.
• Test with diverse records — Do not just test with one record. Try edge cases — records with missing fields, long descriptions, and unusual scenarios.
• Version control — Save prompt templates with version notes so you can track changes over time.
• Ground with Knowledge — For support use cases, always enable Knowledge grounding so the model can reference your articles.
• Use Flex templates for Agentforce — Flex templates are the most versatile and are the type most commonly used in Agentforce actions.

How to Set Up an Agentforce Agent and Deploy It

Now let us build an Agentforce agent from scratch.

Step 1 — Enable Agentforce

1. Navigate to Setup > Agentforce.
2. If Agentforce is not yet enabled, click Enable Agentforce.
3. Review and accept the Einstein Generative AI terms of service if prompted.
4. Wait for provisioning to complete.

Step 2 — Create the Agent

1. In Setup > Agentforce > Agents, click New Agent.

2. Configure the agent basics:

   • Name: Customer Service Agent

   • API Name: Customer_Service_Agent

   • Description: An AI agent that handles customer service inquiries including order status, case creation, and FAQs.

   • Role Description: This is the persona instruction that shapes how the agent behaves. Example:

     You are a helpful, professional customer service agent for our
     company. You are friendly, empathetic, and solution-oriented.
     You always address the customer by name when possible. If you
     cannot resolve an issue, you escalate to a human agent
     gracefully.

3. Click Save.

Step 3 — Define Topics

Topics organize the agent’s capabilities into logical groups. For each topic, you define instructions and associate actions.

1. On the Agent detail page, navigate to the Topics section.

2. Click New Topic.

3. Configure the first topic:

   • Name: Order Inquiries

   • Description: Handles questions about order status, shipping, tracking, and delivery.

   • Classification Description: Customer is asking about an order, shipment, tracking number, delivery date, or package status.

   • Instructions:

     When the customer asks about an order:
     1. Ask for their order number if they have not provided it.
     2. Look up the order using the "Get Order Status" action.
     3. Provide the current status, estimated delivery date, and
        tracking number if available.
     4. If the order is delayed, apologize and offer to escalate.

4. Click Save.

5. Repeat for additional topics:

   • Case Management — Creating, updating, and checking case status.
   • Account Information — Address changes, contact updates.
   • FAQ — Answering common questions using Knowledge articles.
   • Escalation — Handing off to a human agent.

Step 4 — Create Actions

Actions are the tools the agent uses to get things done. Each action is backed by a Flow, Apex class, Prompt Template, or MuleSoft API.

Action Example 1: Get Order Status (Flow-based)

1. First, create a Flow:

   • Create an Autolaunched Flow named Get_Order_Status.
   • Add an input variable: OrderNumber (Text).
   • Add a Get Records element to query the Order object where Order Number equals the input.
   • Add output variables for Status, Tracking_Number, and Estimated_Delivery_Date.
   • Save and activate the flow.

2. Back in the Agent configuration, navigate to the Order Inquiries topic.

3. Click Add Action.

4. Select Flow as the action type.

5. Choose the Get_Order_Status flow.

6. Map the input: OrderNumber — describe it as “The customer’s order number.”

7. Map the outputs: Describe each output so the agent knows how to use the data in its response.

8. Click Save.

Action Example 2: Create a Case (Flow-based)

1. Create an Autolaunched Flow named Create_Support_Case.

   • Input variables: Subject (Text), Description (Text), Priority (Text), ContactId (Text).
   • Add a Create Records element to create a Case record with the input values.
   • Output variable: CaseNumber (the auto-number of the newly created case).
   • Save and activate.

2. Add this flow as an action under the Case Management topic.

3. Map all inputs and outputs with clear descriptions.

Action Example 3: Search Knowledge (Prompt Template-based)

1. Create a Flex Prompt Template named Search_Knowledge_Articles.
   • In the prompt, instruct the model to search for and summarize relevant Knowledge articles based on the customer’s question.
   • Enable Knowledge grounding.
2. Add this prompt template as an action under the FAQ topic.

Step 5 — Configure Guardrails

Guardrails prevent the agent from behaving in undesirable ways.

1. On the Agent detail page, navigate to the Guardrails section.
2. Define guardrails at the agent level (applies to all topics) and at the topic level (applies only to specific topics).

Common guardrails:

Step 6 — Test the Agent

1. On the Agent detail page, click Open in Agent Builder (or Test in newer releases).
2. The testing interface provides a chat window where you can interact with the agent as a customer would.
3. Enter test messages and verify:
   • The agent correctly classifies the topic.
   • The agent follows the topic instructions.
   • The agent calls the correct actions.
   • The agent uses the action results in its response.
   • Guardrails are enforced.
4. Use the conversation log panel to see the reasoning chain — which topic was matched, which actions were called, what data was retrieved, and what guardrails were checked.
5. Iterate on instructions, action descriptions, and guardrails until the agent performs consistently.

Step 7 — Deploy the Agent

Once testing is complete, deploy the agent to a live channel.

Deploying to Service Console

1. Navigate to Setup > Agentforce > Channels.
2. Click New Channel Deployment.
3. Select Service Console as the channel.
4. Choose the agent to deploy.
5. Configure the deployment settings:
   • Placement: Where the agent appears in the console (sidebar, tab, popup).
   • Availability: Which user profiles or permission sets can interact with the agent.
   • Fallback: What happens when the agent cannot resolve the issue (transfer to queue, create a case, etc.).
6. Click Deploy.

Deploying to Experience Cloud

1. In Experience Builder, open your site.
2. Drag the Agentforce component onto a page.
3. Configure the component properties:
   • Select the agent.
   • Set the chat window style (embedded, floating, full-page).
   • Configure pre-chat form fields if needed.
4. Publish the site.

Customers visiting your Experience Cloud site will now interact with the Agentforce agent.

Setting Up an Agentforce Agent with Omni-Channel to Manage Cases

One of the most powerful deployment patterns is connecting an Agentforce agent to Omni-Channel so the agent acts as a first responder for incoming cases. The agent handles what it can and seamlessly transfers to a human agent when needed.

Architecture Overview

The flow works like this:

1. A customer submits a case (via Web-to-Case, Email-to-Case, or messaging).
2. Omni-Channel routes the case to the Agentforce agent as the first-line responder.
3. The agent engages the customer, retrieves context from CRM and Data Cloud, and attempts to resolve the issue.
4. If the agent resolves the issue, it updates the case and closes the conversation.
5. If the agent cannot resolve the issue, it transfers the conversation to a human agent queue through Omni-Channel.
6. The human agent receives the full conversation history and continues where the AI left off.

Step-by-Step Configuration

Part A — Configure Omni-Channel (if not already set up)

1. Navigate to Setup > Omni-Channel > Omni-Channel Settings.
2. Enable Omni-Channel if not already enabled.
3. Create a Service Channel for Cases (if one does not already exist):
   • Navigate to Setup > Omni-Channel > Service Channels.
   • Click New.
   • Name: Case_Channel.
   • Salesforce Object: Case.
   • Save.
4. Create Routing Configurations:
   • Navigate to Setup > Omni-Channel > Routing Configurations.
   • Create a routing configuration for the Agentforce queue:
     • Name: Agentforce_Routing.
     • Routing Model: Most Available (or Least Active).
     • Units of Capacity: 1 per case.
     • Save.
   • Create a second routing configuration for the human agent queue:
     • Name: Human_Agent_Routing.
     • Routing Model: Most Available.
     • Units of Capacity: 1 per case.
     • Save.
5. Create Queues:
   • Navigate to Setup > Queues.
   • Create a queue for the Agentforce agent:
     • Name: Agentforce_Queue.
     • Supported Objects: Case.
     • Routing Configuration: Agentforce_Routing.
     • Queue Members: Add the Agentforce service user (the user that the agent runs as).
   • Create a queue for human agents:
     • Name: Human_Support_Queue.
     • Supported Objects: Case.
     • Routing Configuration: Human_Agent_Routing.
     • Queue Members: Add your human support agents.
6. Create Presence Statuses:
   • Navigate to Setup > Omni-Channel > Presence Statuses.
   • Create or verify a status for online agents (e.g., Available - Support).
   • Assign the presence status to the appropriate profiles or permission sets.

Part B — Configure Case Assignment to Route to the Agent First

1. Navigate to Setup > Case Assignment Rules.
2. Create or modify an assignment rule:
   • Rule Name: Route_to_Agentforce_First.
   • Add a rule entry:
     • Criteria: All new cases (or specific criteria like Origin = “Web” or “Email”).
     • Assign to: Agentforce_Queue.
3. Make this the active assignment rule.

Now all matching cases will land in the Agentforce queue first.

Part C — Configure the Agentforce Agent for Omni-Channel

1. In the Agentforce agent configuration, add an Escalation topic (if you have not already):

   • Name: Escalation

   • Classification Description: The customer wants to speak with a human agent, or the AI agent cannot resolve the issue.

   • Instructions:

     When you need to escalate:
     1. Summarize the conversation so far for the human agent.
     2. Transfer the case to the Human_Support_Queue.
     3. Inform the customer that they are being connected to a
        human agent and provide an estimated wait time if available.

2. Create the escalation action:

   • Build an Autolaunched Flow named Transfer_to_Human_Agent:
     • Input: CaseId (Text).
     • Use an Update Records element to change the Case Owner to Human_Support_Queue.
     • This triggers Omni-Channel to re-route the case to the human queue.
   • Add this flow as an action under the Escalation topic.

3. Add an escalation guardrail at the agent level:

   If the customer explicitly asks to speak with a human, immediately
   transfer to the Human_Support_Queue without further attempts to
   resolve the issue. If you have attempted to resolve the issue twice
   and the customer is not satisfied, transfer to the Human_Support_Queue.

Part D — Configure the Service Console for Handoff

1. In App Builder, open the Service Console Lightning app.
2. Ensure the Omni-Channel Utility is added to the utility bar.
3. Verify that the Agentforce sidebar or component is present on the Case record page.
4. When a human agent accepts a transferred case, they should see:
   • The full conversation history in the Activity Timeline or Chat Transcript related list.
   • The agent’s summary of the issue (generated during the escalation action).
   • All relevant case details populated by the Agentforce agent.

Part E — Test the End-to-End Flow

1. Submit a test case via Web-to-Case or Email-to-Case.
2. Verify Omni-Channel routing — The case should appear in the Agentforce queue and be picked up by the agent.
3. Interact as a customer — Use the testing interface or an Experience Cloud site to chat with the agent.
4. Test resolution — Submit a question the agent can answer. Verify the case is updated and the conversation is closed.
5. Test escalation — Submit a question the agent cannot answer, or explicitly ask for a human. Verify:
   • The case owner changes to the Human Support Queue.
   • Omni-Channel routes the case to an available human agent.
   • The human agent sees the full conversation history.
6. Test guardrails — Try to get the agent to do something outside its scope. Verify it refuses or escalates.

Omni-Channel + Agentforce Best Practices

Comparing Data Cloud and Agentforce Roles

Data Cloud provides the unified data that makes Agentforce agents accurate and context-aware. Without Data Cloud, agents would be limited to the data in standard CRM objects. With Data Cloud, agents can access a 360-degree customer profile that spans every touchpoint.

Section Notes

This post covered two major features that represent the future direction of the Salesforce platform.

Data Cloud Key Takeaways

• Data Cloud is a hyperscale data platform that ingests, harmonizes, and activates customer data from any source.
• The five-layer architecture (Ingestion, Mapping, Identity Resolution, Insights/Segmentation, Activation) is the conceptual framework to keep in mind.
• Data streams bring data in. DLOs are raw staging tables. DMOs are the canonical model. Identity resolution creates unified profiles. Segments group profiles. Activations push data out.
• Start simple — enable the CRM connector bundles, map a few objects, run identity resolution, and explore unified profiles before adding external sources.

Agentforce Key Takeaways

• Agentforce lets you build autonomous AI agents that reason, retrieve data, and take action — all configured declaratively.
• Agents are organized into topics, each with its own instructions and actions.
• Actions are backed by Flows, Apex, Prompt Templates, or MuleSoft APIs.
• Guardrails are critical — they define the boundaries of what the agent can and cannot do.
• Prompt Builder is the tool for creating reusable, data-grounded prompt templates.
• The Omni-Channel integration lets agents act as first responders with seamless handoff to human agents.

Admin Exam Notes

For the Salesforce Admin certification and the AI Associate certification, keep these points in mind:

What We Covered in Topic 2

This post closes out Topic 2 of our Salesforce series. Over the course of Parts 15 through 36, we covered:

• Sales Cloud — Leads, Opportunities, Products, Quotes, Forecasting, Territory Management
• Service Cloud — Cases, Entitlements, Omni-Channel, Knowledge, Field Service
• Experience Cloud — Sites, Templates, Builder, Authentication, CMS
• Flows — Flow Types, Elements, Resources, Debugging, Best Practices
• Approval Processes — Steps, Approvers, Actions
• Retired Automation — Workflow Rules, Process Builder
• Reports and Dashboards — Report Types, Filters, Groupings, Dashboard Components
• List Views — Filters, Charts, Kanban
• Lightning App Builder — Record Pages, Home Pages, App Pages, Dynamic Forms
• Quick Actions, Buttons, and Links
• Paths — Guidance for Success
• Chatter — Posts, Groups, Streams, Topics
• Topics — Topic Assignment, Suggested Topics
• Formulas, Lookup Filters, Custom Labels, and Validation Rules
• Apps and Tabs — Custom Apps, Lightning Apps, Tab Visibility
• Data Cloud — Data Streams, Data Model, Identity Resolution, Segmentation, Activation
• Agentforce — Agents, Topics, Actions, Guardrails, Prompt Builder, Omni-Channel Integration

That is a comprehensive foundation for any Salesforce admin. With Topic 2 complete, you have the knowledge to configure and manage the major products and features on the platform.

Summary

In this final post of Topic 2, we covered:

• Data Cloud — the hyperscale data platform that unifies customer data from every source through data streams, data model objects, identity resolution, calculated insights, segmentation, and activation.
• Data Cloud setup — a step-by-step walkthrough from enabling Data Cloud through creating and activating a segment.
• Agentforce — Salesforce’s platform for building autonomous AI agents with topics, actions, guardrails, and multi-channel deployment.
• Prompt Builder — the declarative tool for creating reusable, data-grounded prompt templates with merge fields and Knowledge grounding.
• Agent setup — a complete walkthrough from enabling Agentforce through deploying a working agent.
• Omni-Channel integration — configuring an Agentforce agent as a first responder that handles cases and escalates to human agents when needed.

Data Cloud and Agentforce together represent a paradigm shift in what admins can accomplish without code. A unified data foundation combined with AI-powered agents means faster service, more personalized experiences, and a more efficient support operation.

Next up — Part 37: Introduction to Apex — We begin Topic 3: “The Complete Guide to Apex.” We will start with the fundamentals of the Apex programming language — syntax, data types, variables, operators, control structures, and your first Apex class. If you have been working purely on the declarative side until now, this is where we cross into code. See you there.