Zero to Hero Salesforce Agentforce 3.0 | 2026

Explore the agent builder and templates in Salesforce Agent Force, deactivate the default agent, and configure topics and actions for customizable service templates.

Build an IT helpdesk agent that initiates conversations, collects employee ID, issue details, and priority, creates a Salesforce case, returns case number, with 24x7 availability, 85% faster, and 100% accuracy.

Learn how employee agents enable internal productivity within Salesforce by handling HR support, IT helpdesk, and operations, with user-permission aware, context-aware filtering and human hand-off.

Create a new it support agent in Agent Force by selecting the employee type, configuring name and role, and assigning topics, data sources, and context variables for testing and debugging.

Create a flow action for agent force by an auto launched flow that queries a contact by email, creates a case with priority and description, and returns the case number.

Create topics for your Salesforce agent from assets or the agent builder, with ai-generated name, classification, and instructions. Map topics to actions like case creation and ticketing, including contact linking.

Assign users to the ai agent in Salesforce, create and apply a permission set, and test with a restricted profile user to verify agent access and case creation.

Explore the theory behind the employee agent, the magic that makes it unique, and its trust layer as we introduce this section.

Atlas reasoning engine powers Agent Force by integrating CRM data and Data Cloud, enforcing security boundaries, and using reinforcement learning to refine structured reasoning and actions.

Explore the Einstein trust layer in Salesforce, covering data privacy, workspace isolation, transparency, access control, and toxicity detection, plus data masking, zero data retention, and audit trails.

Explore dynamic grounding in Salesforce Agentforce, where live data from the data layer is retrieved and used to draft emails via Atlas engine and retrieval augmented generation, respecting permissions.

Enable data masking in data cloud, configure pattern-based and field-level masking via the Einstein trust layer, and apply data sensitivity and compliance categories to secure PII and credit card data.

Explore toxicity detection in the Einstein test layer, including prompt injections and enabling prompt and response toxicity checks, with audit trails and toxicity scoring (zero = unsafe, one = safe).

Enable enhanced event logs to review agent behavior and capture conversations, then generate audit and feedback reports in Data Cloud to assess toxicity scores, masking, and trust layer safety checks.

Explore how prompt templates power Agent Force to generate personalized emails and summaries from leads and products. Learn to configure prompts, call Apex classes and Flows, and store results.

Explore Salesforce prompt templates in the prompt builder, including anomaly analysis and anomaly detection, case details, knowledge article summary, and Slack channel summarization for automated insights.

Learn to build and configure prompt templates in the prompt builder, covering template settings, resolve prompts, model and retriever options, and preview data.

Create a knowledge answers prompt template to generate an executive account summary using web retrieval, including location, headquarters, size, industry, and business details, with data missing when unavailable.

Use the agent force data library to upload a pdf, create a prompt template, and configure a custom retriever to answer questions from the document.

Learn to create a field generation prompt template in Salesforce using the prompt builder, test prompts with OpenAI GPT 3.5, and integrate with the UI via Einstein generative AI.

Use flows and Apex in Salesforce prompt templates to call external APIs and return a string. Pass city data from accounts to drive the weather API and format results.

Call Apex from the prompt builder to generate account data, using an invocable class and related entry to pass the account and return contact names via a field completion template.

Learn to build a record summary with the prompt builder, enabling Copilot to summarize accounts with customized data, callouts, and Apex or Flow resources.

Discover how to use flex templates to handle multiple resources and partners on opportunities, configure accounts like system integrator, developer, and supplier, and generate HTML responses via flows and apex.

Learn how to call a flex template from a Lightning web component by wiring a prompt template through a link component, via prompt builder, returning HTML to display in UI.

Explore how the email content prompt template in Salesforce uses Einstein for sales and language models to generate drafts, with templates and previews for contacts or leads.

Explore how to integrate an apex class into a send email prompt template, returning order and product details through a structured request and response for Einstein GPT sales email.

Learn to call Salesforce prompt templates from external apps via APIs using postman, manage authentication, provide sender, recipient, and product inputs, and handle output from templates.

Explore how service agents extend customer support outside Salesforce to websites, apps, SMS, WhatsApp, and Teams, with security and configuration shaping access.

Explore the use case for customer shopping assistance using a service agent in Experience Cloud, configuring product categories and leveraging prompts to extract keywords, fetch data, and display results.

Learn to fetch product data with apex by building a product finder service and an invocable method, including wrapper inputs, dynamic where clauses, and fuzzy matching.

Create and assign a permission set for the service agent user, granting object and Apex access, then enable and review debug logs to verify queries and event tracking.

Deploy the agent to Experience Cloud by enabling omnichannel, configuring routing and queues, building an omnichannel flow, and setting up a messaging channel and Experience Cloud site for live chats.

Set up messaging channels and presence statuses, configure access with a permission set, enable omnichannel in the service console, and implement outbound channel flow to route to a human agent.

Build and activate an omnichannel flow to route chats to a human agent by linking a created case to the messaging session using the current contact.

Connect multiple systems with Salesforce Data Cloud to create a 360-degree view of the customer by centralizing data from CRM, e-commerce, and marketing sources in real time.

Explore how Salesforce data cloud unifies data from multiple sources, uses connectors and APIs to ingest, harmonizes and resolves identities, and enables insights and activation across marketing and external platforms.

Learn how data sources feed data streams (DSO) into data lake objects and DMO, then harmonize, resolve identities, create unified profiles, and activate segments for targeted analytics.

Upload a file to data cloud, convert a data lake object into a data model object, and use a custom retriever with prompt templates in agent force.

Upload Zara customer data to Data Cloud as CSV, create a data lake object, map to a custom data model object, and verify in the data explorer for agent access.

Explore how the search index and vectorization power agent force searches across data cloud, using dense vectors and vector search and hybrid search for fast, accurate data retrieval.

Create a search index for Zara customer data by configuring data model objects, generating data chunks, applying vectorization, and choosing hybrid or vector search until the index is ready.

Create a custom retriever in Einstein Studio to bridge Data Cloud and Agent Force, configure search index, filters, and shared attributes, activate the retriever, and demo its use.

Create a prompt template to call a custom retriever for Zara customer data, using free text user questions to fetch the customer lifetime value from a DMOZ indexed data source.

Define bring your own model in Salesforce agent force and distinguish predictive from foundational models while exploring cost, compliance, and innovation benefits.

Tune your own models by adjusting temperature, frequency penalty, and presence penalty to shape outputs, choosing low or high values for precise, data-driven responses or more human-like conversations.

Create and deploy a custom predictive model in Salesforce Einstein Studio by ingesting data streams, defining a primary key, training a binary classifier, evaluating metrics, and activating the model.

Learn how to enable Einstein for developers in VS Code, ask Einstein to write Apex classes and methods, check code coverage, and generate predictions, using the Salesforce extension package.

Learn to use Einstein for developers in VSCode to generate an Apex method that fetches account contacts, enable caching for LWC, and deploy to Salesforce.

Generate a test class for Apex with Einstein, fix with sharing and security enforce PMD warnings, and review data setup and test quality before deployment.

Explore Einstein Copilot’s predictive code suggestions for Salesforce test data generation, creating accounts and contacts automatically. Review and adjust Copilot outputs, as prompts may require manual refinements before deployment.

Learn to build a javascript method that imports an apex method, passes account id, stores the context record, and wires the contact data in a lightning web component with Einstein.