Agentforce Testing Middle allows groups to check Agentforce utilizing synthetically generated knowledge, making certain correct responses and actions — with full monitoring of utilization and suggestions
Salesforce (NYSE: CRM), the AI CRM, introduced agentic lifecycle administration instruments to automate Agentforce testing, prototype brokers in safe Sandbox environments, and transparently handle utilization at scale.
AI agents are a brand new paradigm in software program. They’re clever methods that may purpose and act on behalf of consumers and staff. However to understand their full potential, brokers have to be examined and configured with out disrupting reside manufacturing environments. This new toolchain — the primary of its form within the business — will allow groups to check, deploy, and monitor AI brokers with Agentforce at scale, with confidence, enabling each enterprise to change into “agent-first.”
“Agentforce helps companies create a limitless workforce. To ship this worth quick, CIOs want new instruments for testing and monitoring agentic methods,” stated Adam Evans, EVP and GM for Salesforce AI Platform. “Salesforce launched the idea of Software Lifecycle Administration again in 2006 with Pressure.com. This new class of Agentic Lifecycle Administration requires distinctive instruments, and Salesforce is assembly the second once more with Agentforce Testing Middle, which is able to assist firms roll out trusted AI brokers with no-code instruments for testing, deploying, and monitoring in a safe, repeatable manner.”
Innovation in motion: Different distributors lack the required capabilities for patrons to run the suitable checks on their AI earlier than deploying, which may result in hallucinations, inaccurate outcomes, and substandard buyer experiences. Agentforce Testing Middle — constructed on the enterprise-grade Salesforce Platform and built-in with Data Cloud — allows each group to simply take a look at and monitor AI brokers in order that they’ll deploy with confidence.
New capabilities embrace:
- AI-generated checks for Agentforce: Groups constructing with Agentforce must precisely take a look at all the alternative ways a buyer could pose a query or work together with an agent. Along with Agent Builder, which incorporates a Plan Tracer for investigating the reasoning technique of an agent, the brand new Agentforce Testing Middle allows groups to check subject and motion choice at scale. Utilizing pure language directions, Testing Middle can auto-generate tons of of artificial interactions — comparable to requests a buyer could make when participating with Agentforce Service Agent — then take a look at them in parallel to see how regularly they end in the appropriate end result. Groups can then use the take a look at knowledge to refine directions so the anticipated subject is extra regularly chosen, enhancing the top buyer expertise.
- Sandboxes for Agentforce and Knowledge Cloud: Groups seeking to take a look at Agentforce want to take action in protected, remoted environments. Usually accessible at the moment, Salesforce Sandboxes — mirror pictures of your manufacturing org’s knowledge and configurations — now help each Knowledge Cloud and Agentforce. By replicating the org’s knowledge and metadata right into a risk-free surroundings, growth groups can quickly assemble their unstructured knowledge basis and rigorously prototype Agentforce with out worry of disrupting the enterprise. Now groups can carry out UAT (Consumer Acceptance Testing) with an preliminary set of customers to make sure that Agentforce performs the duties that it’s meant to perform, then migrate these adjustments to manufacturing utilizing acquainted instruments comparable to Change Units, DevOps Middle, and the Salesforce CLI that now help Knowledge Cloud and Agentforce.
- Monitoring and observability for Agentforce: With the overall availability of Knowledge Cloud Sandboxes, the total Einstein Trust Layer might be examined in a safe, pre-production surroundings, enabling speedy configuration of Agentforce brokers and Immediate Templates. With the Einstein Belief Layer’s audit path and suggestions retailer in sandboxes, groups can construct a closed loop for AI testing — iterating on prompts and actions primarily based on person suggestions. And as soon as Agentforce is reside in manufacturing, new capabilities for granular insights into adoption and accuracy change into accessible by means of Agentforce Analytics and Utterance Evaluation – new observability options constructed natively on Knowledge Cloud for steady iteration whereas shifting by means of the Agentforce lifecycle.
- Clear utilization monitoring in Digital Pockets: Knowledge Cloud Sandbox and Agentforce utilization is metered in Digital Pockets, offering prospects with full visibility into their consumption throughout the AI growth lifecycle. New enhancements present granular insights into what options devour credit, in order that groups can uncover new tendencies round utilization as they scale. And since Digital Pockets is built-in into the Salesforce Platform, groups can create automations to alert admins if, for instance, utilization exceeds a specific threshold.
Join the free insideAI Information newsletter.
Be part of us on Twitter: https://twitter.com/InsideBigData1
Be part of us on LinkedIn: https://www.linkedin.com/company/insideainews/
Be part of us on Fb: https://www.facebook.com/insideAINEWSNOW
Examine us out on YouTube!
