Salesforce Inc. today launched Agentforce 2.5, the next iteration of its platform for artificial intelligence agents, which allows agents to act proactively based on changes, function autonomously and interact with users more efficiently.
Additionally, Salesforce unveiled AgentforceDX, a toolkit for both developers and business users. The toolset enables faster building, customization, testing and deployment of Agentforce agents. It also offers advanced analytics and monitoring capabilities for easy debugging and performance optimization.
“Companies today have more work than workers, and Agentforce is stepping in to fill the gap,” said Adam Evans, executive vice president and general manager of Salesforce’s AI platform. “By extending digital labor beyond customer relationship management, we’re making it easier than ever for businesses to embed agentic AI into any workflow or application to handle routine tasks.”
AI agents have become a significant industry trend, marking a transformative shift in labor practices, according to Salesforce. Agentic AI refers to systems capable of taking independent actions, making decisions, and utilizing various tools without constant human oversight. In contrast to traditional chatbots, which require specific, repeated prompts, AI agents dynamically respond to real-time data and can automatically execute actions, integrating with industry tools as needed.
Until now, Agentforce agents were triggered from chat interfaces and acted when humans told them to take action. With Agentforce 2.5, agents will operate in the background and can be tied to events using the Agentforce application programming interface that will allow them to allow software to talk to and trigger agents to take action.
That means a developer can build an agent that springs into action when a particular event happens. For example, an AI agent working at a warehouse could be triggered by an alert from data analysis that a particular product is currently low and upcoming orders will require certain amounts available. The agent can then retrieve predictive data about how much to order, and the best prices, and formulate a plan to fulfill the future orders at the lowest possible cost before the surge happens.
Similarly, AI agents could be triggered by customers from Slack events, and pass along Slack conversation context to Agentforce without user-initiated action. That would allow the AI agent to pull together conversation notes about meetings, projects or other conversations that fit particular topics to keep teams up to speed without them asking, or even helpfully put together charts and other elements before they ask.
AgentforceDX, a low-code and pro-code AI agent builder
With the new AgentforceDX, business users and expert users will find a new way to use the company’s Agent Builder tool to get exactly what they want out of Agentforce agents, the company said.
It launches with Agentforce Developer Edition, a new free environment linked with Agentforce and Data Cloud that includes a data space, 10 gigabytes of access and 150 large language model generations per hour.
Customers can get up to speed quickly by asking the AI to assist them in building their first agent and telling it what they want using natural language. By instructing the Agent Builder, they can quickly customize their agent to fit their industry and configure it with best practices.
AI assist can also help teams troubleshoot common problems with prompting, such as accidentally contradicting themselves in long prompts. That includes guidance on how to improve the underlying topic and instructions.
After building and customizing the agent and before deploying, customers can test their Agentforce AI agent at large scale – with as many iterations as they want – in a testing center. It will automatically generate as many test cases as they want and deploy them into a sandbox so they can evaluate scenarios to make sure it is adhering to guardrails such as faithfulness and relevance.
Professional, or pro-code, developers can create, update and test Agentforce agents in their preferred environment including the command line and Visual Studio. Every aspect of Agentforce can be configured, created and updated with pro code tools, and test cases can be run for these configurations, the company said.
Once the agent is released to the public, users get access to a dashboard called Agentforce Interaction Explorer that provides detailed reporting and analytics about how agents are performing. It also offers a broad holistic view that includes general trends down to individual sessions. That allows AI engineers to drill down into conversations to understand the logical progression and reasoning of individual AI agents to help pull them out and refine their prompts in the testing center.
The release of this development toolset follows the announcement Monday of AgentExchange, an AI agent skills and templates marketplace launched with more than 200 initial integration partners earlier this week. Developers can use “skills” taken from partners in the marketplace and integrate them into their agents to accelerate their development, or choose from industry templates to jumpstart the process of building and deploying.
Image: News/Microsoft Designer
Your vote of support is important to us and it helps us keep the content FREE.
One click below supports our mission to provide free, deep, and relevant content.
Join our community on YouTube
Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.
THANK YOU
