Salesforce is updating its agentic development platform, infusing AI agents with new autonomous capabilities. With the release of Agentforce 2dx, created bots are moving away from the confines of a chat interface and can now be embeddable in any application, workflow, or process.
The newest Agentforce update, revealed at Salesforce’s TDX conference, includes new features designed to help organizations tap into the AI-powered digital workforce. It also enables developers to integrate these agents easily into existing data systems, business logic, and user interfaces, reducing potential adoption barriers.
“Companies today have more work than workers, and Agentforce is stepping in to fill the gap,” said Adam Evans, the Executive Vice President and General Manager of Salesforce’s AI platform, in a release. “By extending digital labor beyond CRM, we’re making it easier than ever for businesses to embed agentic AI into any workflow or application. With deep integrations across Salesforce’s digital labor platform, CIOs, IT leaders, and developers can seamlessly build agents and automate work wherever it happens, driving efficiency, fueling innovation, and unlocking new opportunities in the $2 trillion AI agent market.”
Although announced today, not every feature in Agentforce 2dx will be available immediately. Some are launching today, while others could arrive later this fall.
Disclosure: Salesforce has invited me to attend TDX as its guest, paying for my expenses. The company, in no way, dictated the content of this post. These are my words.Transforming Agents from Reactive to Proactive
Salesforce appears to be putting the proverbial pedal to the metal regarding AI agents. This is the second significant platform release in as many months since its October 2024 launch. Since then, the company has sought to redefine the role bots play across every corner of an organization’s business, expanding its reach into Slack. Fast-forward to today, and Salesforce wants to highlight the unrealized full potential agents can play when they can access all the necessary data.
It’s also another hard push by Salesforce to prove to customers that its AI technology can generate real value—something Chief Executive Marc Benioff has chastised rival Microsoft over with its Copilot offering.
With Agentforce 2dx, Salesforce said it’s breaking agents free from the chatbox construct. The tech firm doesn’t view these bots as something you exclusively interact with through a messaging window. Evans once claimed, “We’re at the beginning of a major transformation in which autonomous AI agents will become the new user interface.” The latest platform release is an attempt to bring that vision to life.
“It starts with this idea of being able to do background work where agents can be triggered from any platform event anytime a record is changed [or] new cases come in,” Evans explained during a press conference on Monday. “Anything inside of the unified platform of Salesforce can begin an agent to do work in the background as a data analyst and more. And this change is significant because now, instead of chat, where agents tend to be reactive, Agentforce has this ability to be proactive and, in adding this, we also have more control that enterprises demand with the introduction of variables that allows us to blend deterministic logic with variables and structures with the amazing capabilities of the thinking and unstructured thinking that agents provide altogether.”
Salesforce claimed these proactive agents will dynamically respond to new real-time inputs and changing business needs. It boasts that doing so marks a “major leap in how organizations integrate agents into daily operations.” Instead of resigning teams to hard-coding automation and designating triggers, companies can double down on AI apps that monitor, freeing up resources to pursue other initiatives. AgentExchange, Salesforce’s new agentic marketplace, will play a critical role in making this development.
Subscribe to The AI Economy
New Tools to Build and Deploy Agents
So, what’s new with Agentforce 2dx? Salesforce has packaged a suite of tools that it boasts will help developers “embed Agentforce anywhere” and customize it for any use case. Here are some of the more notable mentions:
AgentforceDX and Developer Edition
This set of no-code and low-code tools is designed to help stand up, test, and deploy Agentforce. It has advanced analytics to monitor, debug, and optimize agent performance in real time. This offering is available with the new Agentforce Developer Edition, a free-to-use development environment for prototyping agents with one data space, 10GB of Data Cloud access, and 150 LLM generations per hour.
Agent Builder with AI Assist
This tool allows builders to configure agents quickly using simple task descriptions. The AI assist feature generates optimized topics and instructions within Agentforce and will provide suggestions on making bots more efficient. This feature will be available in March 2025 as part of Agentforce.
CLI and VS Code Support
Developers can now use Agentforce with the Command Line Interface (CLI) and Visual Studio Code (VS Code). As Salesforce describes it, platform configurations can be created and updated with pro code tools, and test cases can be run against these settings.
MuleSoft for Agentforce Enhancements
Complicated integrations are a critical stumbling block for IT teams when completing AI projects on time. A recent MuleSoft survey shows that nearly a third of these initiatives failed to be delivered on schedule. So, if Salesforce does more to make using AI more like a “plug and play” experience, it will be less work for developers and a win for Salesforce.
To that end, Salesforce is releasing several MuleSoft updates with Agentforce 2dx. The first improvement is the ability to use natural language to create Agentforce topics and actions from the MuleSoft API. Second, MuleSoft’s API Catalog has now reorganized API access from MuleSoft, Salesforce, and Heroku within Agentforce, providing a secure and governed location for administrators to import agent capabilities. Lastly, the MuleSoft Agentforce connector now supports integrating AI agents into existing workflows through APIs and integrations.
This is all available today.
Agentforce Interaction Explorer
If you’re curious about how your AI agent is performing, Agentforce’s Interaction Explorer is here to provide answers. It’s an analytics tool that gauges all performance levels from monitoring high-level trends to individual session interactions.
The Agentforce Interaction Explorer is a tool Evans is particularly excited about. “Our customers are giving us feedback, and we’re able to understand what agents are doing, how they impact the business, and how customers are liking it. How do you take these learnings and know where to continue to invest and continually improve? This is a huge investment in that area,” he said.
The Salesforce executive went on to outline three levels of this tool, with the top level being an aggregate that captures the agent’s “inner monologue” as it does the work, tags and flags key moments, and surfaces insights for analysis in tools like Tableau. In the second layer, users drill down into flagged interactions, identify patterns, and are shown AI-generated recommendations for improving agent behavior. Lastly, the third layer is where you have the individual interaction and can review things like latency, what actions were called, the parameters, and other similar kinds of event logging. Developers can use real-world data to refine AI performance through Agentforce’s testing center.
The Agentforce Interaction Explorer will be available in Fall 2025.
Agentforce Testing Center
Speaking of the testing center, starting in April 2025, it will support using AI to test Agentforce configurations across multiple scenarios. Developers can automatically create and run test cases within their Salesforce developer environment and review them to see if the configurations match the guidelines for accuracy and relevance.
In addition to these updates, Salesforce has also made Agentforce improvements regarding rich content, custom response actions for Lightning Web Components, embedding no-code automation in Slack, and more.
Featured Image: Salesforce Agentforce signage showcased at AWS re:Invent on December 3, 2024. Photo by Ken Yeung.
Leave a Reply
You must be logged in to post a comment.