--- name: agentforce-2025 description: Salesforce Agentforce AI agents and autonomous automation (2025-2026). PROACTIVELY activate for: (1) building Agentforce agents and topics, (2) Agent Builder configuration, (3) Atlas Reasoning Engine usage, (4) Agentforce action setup (Apex, Flow, Prompt Templates), (5) topic-based routing and instructions, (6) Agentforce Service Agent vs Sales Agent vs Custom, (7) Einstein Trust Layer (data masking, toxicity detection), (8) prompt engineering for grounded responses, (9) Agentforce testing and Test Center, (10) deploying agents across orgs. Provides: agent design patterns, topic and action templates, prompt engineering recipes, Trust Layer configuration, and deployment workflow. --- ## 🚨 CRITICAL GUIDELINES ### Windows File Path Requirements **MANDATORY: Always Use Backslashes on Windows for File Paths** When using Edit or Write tools on Windows, you MUST use backslashes (`\`) in file paths, NOT forward slashes (`/`). **Examples:** - ❌ WRONG: `D:/repos/project/file.tsx` - ✅ CORRECT: `D:\repos\project\file.tsx` This applies to: - Edit tool file_path parameter - Write tool file_path parameter - All file operations on Windows systems ### Documentation Guidelines **NEVER create new documentation files unless explicitly requested by the user.** - **Priority**: Update existing README.md files rather than creating new documentation - **Repository cleanliness**: Keep repository root clean - only README.md unless user requests otherwise - **Style**: Documentation should be concise, direct, and professional - avoid AI-generated tone - **User preference**: Only create additional .md files when user specifically asks for documentation --- # Agentforce: AI Agents for Salesforce (2025) ## What is Agentforce? Agentforce is Salesforce's enterprise AI agent platform that enables autonomous, proactive applications to execute specialized tasks for employees and customers. Agentforce agents use large language models (LLMs) with the Atlas Reasoning Engine to analyze context, reason through decisions, and take action autonomously. **Key Distinction**: Agentforce represents the evolution from Einstein Copilot (conversational assistant) to autonomous agents that can complete tasks without human prompting. ## Core Architecture ### Atlas Reasoning Engine The Atlas Reasoning Engine is the brain of Agentforce, enabling agents to: - **Understand**: Analyze full context of customer interactions or automated triggers - **Decide**: Reason through decisions using LLMs and business logic - **Act**: Execute actions autonomously across any system - **Learn**: Improve over time based on outcomes and feedback ### Agent Components ``` ┌─────────────────────────────────────────────────┐ │ Agentforce Agent │ ├─────────────────────────────────────────────────┤ │ 1. Topics (what agent handles) │ │ 2. Actions (what agent can do) │ │ 3. Instructions (how agent behaves) │ │ 4. Channel Integrations (where agent works) │ │ 5. Data Sources (what agent knows) │ └─────────────────────────────────────────────────┘ ``` ## Agentforce 2.0 (GA December 2024) Agentforce 2.0 is the digital labor platform for enterprises, enabling a limitless workforce through AI agents. Key enhancements: - **Pre-built Skills Library**: Rapid agent customization with out-of-the-box capabilities - **Workflow Integrations**: MuleSoft for Flow, MuleSoft API Catalog, Topic Center (Q1 2025) - **Slack Deployment**: Native Slack integration for collaboration agents - **Enhanced RAG**: Improved retrieval augmented generation for accurate responses - **Advanced Reasoning**: More sophisticated Atlas Reasoning Engine capabilities - **Pricing**: $2 per conversation (GA October 25, 2024) ## Building Agents with Agentforce Builder (GA December 2024) ### Step 1: Define Agent Purpose Identify what the agent should accomplish: - **Service Agent**: Handle support cases, answer FAQs, resolve issues (4 new actions in Spring '25) - **Sales Development Agent**: Qualify leads, answer product questions, book meetings - **Personal Shopper Agent**: Recommend products, handle orders, track shipments - **Operations Agent**: Automate approvals, process requests, manage workflows - **Slack Agent**: Proactive notifications and collaboration assistance ### Step 2: Configure Agent Topics Topics define what the agent can help with: ```apex // Example: Service Agent Topics Topic: Password Reset - Intent: User wants to reset password - Required Data: Email, Username - Connected Action: PasswordResetFlow Topic: Order Status - Intent: User wants order status - Required Data: Order Number or Email - Connected Action: GetOrderStatus Topic: Escalate to Human - Intent: User needs human assistance - Required Data: Case context - Connected Action: CreateCase + NotifyAgent ``` ### Step 3: Define Agent Actions Actions are what the agent can execute. These can be: - **Standard Actions**: Pre-built Salesforce actions (create/update records) - **Flow Actions**: Custom Flow automations - **Apex Actions**: Custom Apex invocable methods - **MuleSoft Actions**: API integrations via MuleSoft connectors - **External API Actions**: REST/SOAP callouts **Example Apex Action**: ```apex public class AgentActions { @InvocableMethod(label='Get Order Status' description='Retrieves order status for customer') public static List getOrderStatus(List requests) { List results = new List(); for (OrderRequest req : requests) { Order order = [SELECT Id, Status, EstimatedDelivery__c FROM Order WHERE OrderNumber = :req.orderNumber LIMIT 1]; OrderStatus status = new OrderStatus(); status.orderNumber = req.orderNumber; status.status = order.Status; status.estimatedDelivery = order.EstimatedDelivery__c; results.add(status); } return results; } public class OrderRequest { @InvocableVariable(required=true) public String orderNumber; } public class OrderStatus { @InvocableVariable public String orderNumber; @InvocableVariable public String status; @InvocableVariable public Date estimatedDelivery; } } ``` ### Step 4: Write Agent Instructions Instructions guide the agent's behavior and tone: ``` You are a helpful customer service agent for Acme Corp. Personality: - Friendly, professional, and empathetic - Patient with customers who are frustrated - Proactive in offering solutions Guidelines: - Always greet customers by name if available - Verify customer identity before sharing account information - Offer alternatives if the requested action cannot be completed - Escalate to human agent for: refunds >$500, legal issues, abusive customers - Use simple language, avoid jargon - Provide order numbers and case numbers in responses Security: - Never share: passwords, credit card numbers, SSN - Always verify identity using: email, phone, or account number - Log all interactions for compliance Response Format: - Keep responses under 3 sentences when possible - Use bullet points for multiple items - Include next steps or call-to-action ``` ### Step 5: Connect Data Sources Agentforce agents can access: - **Salesforce Objects**: Standard and custom objects - **Data Cloud**: Unified customer data from all sources - **Knowledge Base**: Salesforce Knowledge articles - **External Systems**: Via APIs and MuleSoft connectors **Data Cloud Integration**: ```apex // Query Data Cloud from Agentforce public class AgentDataCloudActions { @InvocableMethod(label='Get Customer 360 View') public static List getCustomer360(List customerIds) { // Query Data Cloud for unified customer data List results = new List(); for (String customerId : customerIds) { // Data Cloud connector provides unified view DataCloudConnector.QueryRequest req = new DataCloudConnector.QueryRequest(); req.sql = 'SELECT * FROM Unified_Customer WHERE customer_id = \'' + customerId + '\''; DataCloudConnector.QueryResponse res = DataCloudConnector.query(req); Customer360 customer = new Customer360(); customer.customerId = customerId; customer.totalPurchases = (Decimal)res.data.get('total_purchases'); customer.preferredChannel = (String)res.data.get('preferred_channel'); customer.lifetimeValue = (Decimal)res.data.get('lifetime_value'); results.add(customer); } return results; } } ``` ### Step 6: Configure Channels Deploy agents across multiple channels: - **Web Chat**: Embedded on website - **Mobile App**: In Salesforce Mobile or custom apps - **SMS/WhatsApp**: Messaging platforms - **Slack/Teams**: Collaboration tools - **Voice**: Phone support with voice-to-text - **Email**: Email case management ## Agent Types and Use Cases ### 1. Service Agent (Customer Support) **Capabilities**: - Answer FAQs from Knowledge Base - Retrieve order/account status - Process returns and exchanges - Reset passwords and unlock accounts - Create and route cases to specialists - Provide troubleshooting steps **Example Flow**: ``` Customer: "Where is my order #12345?" ↓ Agent: Validates order number ↓ Agent: Queries Order object ↓ Agent: Retrieves tracking information ↓ Agent: "Your order #12345 shipped yesterday and will arrive Thursday. Tracking: UPS 1Z999AA10123456784. Need anything else?" ``` ### 2. Sales Development Agent (SDR) **Capabilities**: - Qualify inbound leads - Answer product questions - Handle objections with sales playbooks - Book meetings with sales reps - Send follow-up emails - Update lead scores based on engagement **Example Flow**: ``` Lead: "Tell me about your enterprise plan" ↓ Agent: Retrieves product information ↓ Agent: Explains features, pricing ↓ Agent: Detects buying intent ↓ Agent: "Would you like to schedule a demo with our sales team?" ↓ Agent: Creates meeting, updates lead status to "Meeting Scheduled" ``` ### 3. Personal Shopper Agent (E-commerce) **Capabilities**: - Recommend products based on preferences - Answer product questions - Check inventory and availability - Process orders and payments - Apply discounts and promotions - Handle cart abandonment ### 4. Operations Agent (Internal Automation) **Capabilities**: - Process employee requests (PTO, equipment) - Automate approvals based on rules - Onboard new employees - Generate reports and insights - Monitor system health - Trigger workflows based on events ## Integrating Agentforce with Platform Events Publish events to trigger Agentforce actions: ```apex // Publish event when order status changes public class OrderEventPublisher { public static void publishOrderUpdate(Id orderId, String newStatus) { OrderStatusChangeEvent__e event = new OrderStatusChangeEvent__e( OrderId__c = orderId, NewStatus__c = newStatus, Timestamp__c = System.now() ); EventBus.publish(event); // Agentforce subscribes to this event // Triggers proactive customer notification } } // Trigger trigger OrderTrigger on Order (after update) { for (Order ord : Trigger.new) { if (ord.Status != Trigger.oldMap.get(ord.Id).Status) { OrderEventPublisher.publishOrderUpdate(ord.Id, ord.Status); } } } ``` **Agentforce Flow** (subscribed to OrderStatusChangeEvent__e): ``` 1. Receive event 2. Query order and customer details 3. Determine notification channel (email, SMS, push) 4. Generate personalized message using LLM 5. Send notification via preferred channel 6. Log interaction in Customer timeline ``` ## Agentforce with External AI Systems Integrate Agentforce with external AI providers: ### OpenAI GPT Integration ```apex public class AgentOpenAIIntegration { @InvocableMethod(label='Generate Response with GPT-4') public static List generateResponse(List requests) { List responses = new List(); for (AIRequest req : requests) { HttpRequest httpReq = new HttpRequest(); httpReq.setEndpoint('callout:OpenAI/v1/chat/completions'); httpReq.setMethod('POST'); httpReq.setHeader('Content-Type', 'application/json'); Map payload = new Map{ 'model' => 'gpt-4', 'messages' => new List

agentforce-2025

Install this skill

Comments