Salesforce Launches Agentforce Testing Center: A Pioneering AI Tool for Managing Autonomous Agent Lifecycles at Scale

On November 20, 2024, Salesforce, the world’s leading AI CRM, introduced innovative agentic lifecycle management tools designed to automate Agentforce testing, prototype agents in secure Sandbox environments, and efficiently manage large-scale usage.

AI agents represent a new frontier in software—intelligent systems capable of reasoning and acting on behalf of both customers and employees. To harness their full potential, agents must be tested and configured without interrupting live production environments. This groundbreaking toolchain—an industry-first—empowers teams to test, deploy, and monitor AI agents with Agentforce at scale, all while ensuring reliability and confidence, enabling enterprises to embrace an “agent-first” approach.

“Agentforce is enabling businesses to create a limitless workforce. To realize this potential swiftly, CIOs need advanced tools for testing and monitoring agentic systems,” said Adam Evans, EVP and GM of Salesforce AI Platform. “Salesforce pioneered Application Lifecycle Management with Force.com in 2006. Now, with the rise of Agentic Lifecycle Management, we’re addressing the need for specialized tools with Agentforce Testing Center, which provides no-code solutions for testing, deploying, and monitoring AI agents in a secure, repeatable way.”

Also Read: Boostr and pureIntegration Announce Strategic Partnership

Innovation in Action: While other vendors lack the necessary capabilities to thoroughly test AI systems before deployment—leading to errors such as hallucinations and inaccurate results—Agentforce Testing Center, built on the enterprise-grade Salesforce Platform and integrated with Data Cloud, ensures that AI agents can be tested and monitored effectively before deployment, giving organizations the confidence to scale.

New Capabilities Include:

• AI-Generated Tests for Agentforce: Teams working with Agentforce need to test various customer interactions. In addition to the Agent Builder, which includes the Plan Tracer for analyzing an agent’s reasoning process, the new Testing Center allows teams to test topic and action selections at scale. By using natural language instructions, the Testing Center can automatically generate hundreds of synthetic customer interactions—such as service inquiries—and test them in parallel to assess how often they result in the desired outcome. This data helps refine instructions to improve accuracy, enhancing the customer experience.
• Sandboxes for Agentforce and Data Cloud: Secure, isolated environments are crucial for testing AI agents. Now, Salesforce Sandboxes, which replicate production data and configurations, support both Data Cloud and Agentforce. This enables teams to quickly build their unstructured data foundation and rigorously prototype Agentforce without disrupting business operations. User Acceptance Testing (UAT) can be conducted to verify agent performance, with changes easily migrated to production using familiar tools like Change Sets, DevOps Center, and Salesforce CLI.
• Monitoring and Observability for Agentforce: With Data Cloud Sandboxes available, the full Einstein Trust Layer can be tested in a secure environment. This enables the rapid configuration of Agentforce agents and Prompt Templates. The Trust Layer’s audit trail and feedback store in sandboxes provide a closed-loop system for AI testing, allowing teams to iterate on prompts based on user feedback. After deployment, new capabilities for detailed insights into adoption and accuracy are provided via Agentforce Analytics and Utterance Analysis, which are integrated into Data Cloud for continuous iteration.
• Transparent Usage Monitoring in Digital Wallet: Agentforce and Data Cloud Sandbox usage are metered in Digital Wallet, offering customers full transparency into their consumption during the AI development lifecycle. New enhancements provide detailed insights into which features consume credits, helping teams identify usage patterns as they scale. Integrated with Salesforce, Digital Wallet also enables the creation of automations to alert admins if usage exceeds defined thresholds.

Customer and Analyst Perspectives:

“With nearly 3 million members, PenFed is dedicated to fostering long-term relationships with our customers,” said Shree Reddy, CIO of PenFed. “Using the power of Data Cloud on the Salesforce Platform, we’re providing member service reps with a personalized view of each member’s journey. The ability to thoroughly test prior to deployment with Data Cloud Sandboxes helps us deliver faster, more accurate support and recommendations, reinforcing our commitment to each member’s financial success.”

“Trust in AI will be key for businesses adopting agent technologies, and that means thoroughly testing autonomous AI agents,” said Keith Kirkpatrick, Research Director at The Futurum Group. “Salesforce’s use of AI to test hundreds of variations in parallel, coupled with these new enhancements and the ability to test in familiar environments, ensures that companies can confidently manage the complexities of AI agent deployment.”