agentforce-2025

What this skill does

## 🚨 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

```text
┌─────────────────────────────────────────────────┐
│              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)

## Agentforce Builder Workflow

Detailed Agentforce Builder workflow — agent creation, topics, instructions, actions, grounding, testing, publishing, Data Cloud integration, trust / guardrails, and deployment mechanics — lives in `references/agentforce-builder.md`. Load that reference when building or configuring agents rather than just selecting an architecture.

## 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**:
```text
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**:
```text
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):
```text
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<String> generateResponse(List<AIRequest> requests) {
        List<String> responses = new List<String>();

        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<String, Object> payload = new Map<String, Object>{
                'model' => 'gpt-4',
                'messages' => new List<Object>{
                    new Map<String, String>{
                        'role' => 'system',
                        'content' => req.systemPrompt
                    },
                    new Map<String, String>{
                        'role' => 'user',
                        'content' => req.userMessage
                    }
                },
                'temperature' => 0.7,
                'max_tokens' => 500
            };

            httpReq.setBody(JSON.serialize(payload));

            Http http = new Http();
            HttpResponse httpRes = http.send(httpReq);

            if (httpRes.getStatusCode() == 200) {
                Map<String, Object> result = (Map<String, Object>)JSON.deserializeUntyped(httpRes.getBody());
                List<Object> choices = (List<Object>)result.get('choices');
                Map<String, Object> choice = (Map<String, Object>)choices[0];
                Map<St