Learn About Agentforce Testing Center

Key Features

• Preview agent topic classification for test utterances
• Explore action sequences at scale
• Use result data to refine agent instructions
• AI-generated test cases to save time testing
• LLM-as-a-judge evaluation for response quality assessment
• Multi-turn conversation testing with Agent Builder integration

Who Should Use Testing Center?

Agentforce Testing Center is designed for:

• Salesforce Admins - Configure and manage agent testing workflows
• Agent Builders - Test agent performance before deployment
• Agent Managers - Oversee agent quality and performance
• Business Users - Validate agent behavior meets business requirements
• Developers - Pro-code testing and integration with DevOps workflows

Why Use Testing Center?

Without a dedicated testing facility, you risk deploying agents that:

• Provide incorrect or irrelevant information due to knowledge gaps.
• Deliver inconsistent responses due to inefficient instructions.
• Struggle with complex, ambiguous, or even inappropriate user queries such as prompt injection.
• Fail to adapt to changing contexts or conversation history.
• Agent Hallucinations

The Testing Center empowers you to identify and mitigate these risks at scale, ensuring your agents are robust, reliable, and production-ready.

Availability & Prerequisites

Current Availability

• Sandbox Orgs: Automatically enabled for all Agentforce customers
• Production Orgs: Not recommended for enablement, as it can affect production data

Prerequisites

• Agentforce activated in your Sandbox
• Test utterances prepared in CSV template format
• Org must allow file uploads
• System admin permissions for uploading/downloading test data CSV files

Supported Agent Types

• Agentforce Service Agent (ASA) - Full support
• SDR Agent - Full support
• Employee Agent - Full support
• Default Agent - Full support

Detailed Features & Capabilities

Test Creation Methods

1. CSV Upload - Manual test case creation using provided template
2. AI-Generated Test Cases - Automated generation based on agent configuration
3. Knowledge-Based Generation - Create Q&A pairs from Agentforce Data Library content
4. Conversation History Import - Import multi-turn conversations from Agent Builder

Evaluation Metrics

• Topic Classification - Exact match validation
• Action Sequences - Verify correct action execution
• Response Quality - LLM judge scoring (0-5 scale, ≥3 = Pass)
• Text Quality Metrics - Conciseness, completeness, coherence
• Citation Support - Verify knowledge article citations
• Instruction Adherence - Ensure agents follow tone and guidance
• Latency - Response time measurement

Enhanced Debugging

• Run Status Column - Clear error messaging for failed tests
• Error Tooltips - Help triage runtime issues
• Session Tracing - Full conversation logs 
• Enhanced Error Messages - Better visibility into stuck jobs and rate limits

Getting Started

1. If not already enabled, Activate Agentforce in your Sandbox environment
2. Navigate to Setup → Agentforce Testing Center
3. Download the CSV template or try AI generation
4. Create your first test with 5-10 simple utterances
5. Run the test and review results
6. Iterate and expand based on your agent's use cases

Pricing & Consumption

License Requirements

• No additional license needed - Automatically enabled for all Agentforce customers in Sandbox environments

Consumption Tracking

• Track ER consumption in Digital Wallet
• Monitor usage in Testing Center activity logs

Performance & Limits

Current Limits

• Maximum test cases per job: 500
• Jobs per hour: 10 maximum
• Recommended batch size: 20-30 test cases per evaluation for optimal performance
• Test execution time: ~5 seconds per test case on average

Performance Tips

• Keep executions smaller (20-30 test cases) to avoid rate limits
• Create separate test definitions for different scenarios
• Monitor job progress with refresh button
• Consider multiple smaller jobs vs. one large job

Testing Strategy & Best Practices

Comprehensive Test Coverage

Instead of targeting a specific number, focus on coverage across:

• Features: Case creation, status updates, knowledge articles, order tracking, account information
• Scenarios: Multiple matches found, no match found, incomplete info, channel-specific constraints
• Personas: Authenticated users, unauthenticated users, new users, mobile users
• Edge Cases: Incomplete info, unsupported features, prompt injection scenarios

Recommended Approach

• Start with 30-40 test cases in first iteration
• Generate more data until you stop seeing new failure modes
• Mix test types: Features + Scenarios + Personas + Edge cases
• Regular testing: Test throughout development lifecycle, not just at end

Test Data Guidelines

• Utterance column: Cannot be empty (required)
• Other columns: Can have empty values (handled automatically)
• Expected responses: Review AI-generated responses and override as needed
• Multi-language: Currently English only, multilingual support planned

User Experience & Workflow

Test Creation Workflow

1. Access Testing Center from Setup → Agentforce Testing Center
2. Choose method: Upload CSV, AI generation, or Knowledge-based generation
3. Configure evaluation metrics (response quality, instruction adherence, etc.)
4. Run tests and monitor progress
5. Review results and export for analysis
6. Iterate based on findings

Results Analysis

• Pass/Fail indicators with detailed scoring
• Export capabilities in CSV format for external analysis
• Retry capabilities for failed test cases
• Inline citations for knowledge-based responses

Integration with Agent Builder

• Conversation import from Agent Builder testing sessions
• Multi-turn testing using conversation history
• Seamless workflow between building and testing