Salesforce is set to enhance its presence in the AI sector with the introduction of Agentforce—an innovative platform designed to implement generative AI across customer support, service, sales, and marketing domains. This new offering is scheduled to launch on October 25.
At a press briefing on September 12, Salesforce CEO Marc Benioff highlighted that customers using Agentforce will not need to create their own AI models or manually select from different models. He underscored that the AI service is not simply an add-on “copilot” that clutters existing services.
The functionalities previously known as Einstein Copilot, including those integrated into Slack, will now be rebranded as part of Agentforce.
Transforming chatbots into generative conversationalists
Agentforce serves as a deployment platform for generative AI agents capable of autonomously engaging with both current and prospective customers. The system integrates with Salesforce’s Data Cloud, efficiently processing both structured and unstructured data through the Atlas Reasoning Engine—a key component of the generative AI framework.
The Data Cloud’s data lake offers a “Zero Copy” feature, which allows data to be sourced from external systems without duplication. Although Salesforce has not fully disclosed details about the Atlas Reasoning Engine, they explained that it plays a crucial role in:
- Filtering the decisions made by AI agents
- Determining the relevance of data
- Verifying suggested responses against real-world information
Agentforce agents can conduct autonomous conversations with customers, addressing service inquiries, following up on sales leads, or assisting with marketing initiatives. They can interact via voice, phone, or text. During the September 12 event, Salesforce showcased an AI “personality,” complete with its own name and settings, which can engage multimodally across different platforms.
Benioff suggested utilizing Agentforce as a substitute for level 1 customer service representatives, allowing AI agents to handle routine inquiries while escalating more complex issues to level 2 staff, who possess greater expertise.
Agentforce offers a low-code deployment solution, where organizations can integrate a library of agent actions into a user-friendly menu, with some customization possible through natural language prompts. The Prompt Builder service provides the tools to tailor these autonomous agents using data from the Data Cloud. Organizations wishing to deploy their own preferred language models (LLMs) and APIs can do so within the low-code framework.
Agents can be initially configured with specific roles, with additional access to data granted later. Image: Salesforce
Similar to human agents, Agentforce agents will adhere to certain monitoring and safety standards. The existing Salesforce Omni Supervisor feature will empower supervisors to oversee AI agents, identifying potential issues, and ensuring quality control. Salesforce AI CEO Clara Shih noted that this capability allows human managers to intervene if an AI agent provides inappropriate or inaccurate responses.
“Omni Supervisor now allows service managers to oversee their Agentforce agents as well. They can filter by topic and identify trends. Much like training a new human hire, managers can step in when necessary to ensure smooth operations,” Shih explained.
Tackling the challenges of wait times and phone trees
Benioff pointed out that many organizations are seeking to deliver tangible value to their customers amid the current AI landscape. He noted that customers are looking for ways to reclaim their time, framing Agentforce as a viable alternative to long hold times when reaching customer service representatives.
Pricing for Agentforce
Agentforce will be offered at a cost of $2 per conversation. Salesforce has yet to clarify how this pricing compares to human-staffed service interactions.
While Agentforce for Service and Sales will launch on October 25, some features of the Atlas Reasoning Engine are expected to be available in February 2025.
Upcoming Salesforce updates in October
The October release of Salesforce will also introduce additional features, including a Quota Attainment Planning tool and an AI-driven Employee Service solution within ServiceCloud, aimed at enhancing employee support.
