The Evolution of Agentforce: Salesforce’s Vision for Digital Labor Takes Shape with 3.0

Salesforce has just announced Agentforce 3.0, the latest iteration of its flagship AI product. Since its initial launch in September 2024, Agentforce has become a central pillar of Salesforce’s AI strategy. This newest update signifies that Salesforce is actively responding to feedback from its ecosystem, directly addressing concerns that professionals have voiced over the past few months regarding visibility, control, and integration. Agentforce 3.0 introduces features designed to offer users deeper insight into agent performance, greater governance over their deployment, and more seamless integration with existing business tools and workflows.

The introduction of Agentforce 2.0 late last year hinted at Salesforce’s vision of a digital labor workforce. That update brought new pre-built agent skills for platforms like Slack, Tableau, and AppExchange partners, alongside MuleSoft integrations and an upgraded Atlas Reasoning Engine. However, Salesforce professionals have remained skeptical about Agentforce, citing concerns over security, technical debt, pricing, governance, and ROI. In response, Salesforce has shifted its pricing model from a $2-per-conversation fee to Flex Credits and announced the general availability of dedicated Agentforce user licenses and add-ons, demonstrating a clear commitment to driving adoption and making it easier for customers to get started. This move aims to remove friction and scale usage across teams. Agentforce 3.0 further advances this goal by introducing new ways to monitor agents and connect to external enterprise tools, moving Agentforce beyond a mere prototype to an enterprise-grade AI solution.

Agentforce Command Center: Unprecedented Visibility and Control

One of the most significant features of Agentforce 3.0 is the Agentforce Command Center, a new tool for tracking, monitoring, and measuring AI agents. The Command Center offers a suite of capabilities designed to enhance observability and optimization:

• Real-time Agent Health Monitoring: Users can track performance, error rates, and escalations, with immediate alerts flagging any issues.

• Adoption Insights: The Command Center provides an overview of agent usage, identifying what’s working well and where improvements are needed.

• Pattern Recognition and Optimization: It allows for the analysis of interactions and provides AI-driven recommendations to boost agent performance.

• Comprehensive Interaction Tracing: Built-in session tracing, powered by OpenTelemetry and Data Cloud, enables enterprise-wide observability.

• Team-Specific Command Centers: Users can tailor Command Centers for individual teams, viewing AI and human activity side-by-side, starting with Service Cloud wallboards.

• Faster Development and Testing: The Command Center facilitates agent generation using natural language and allows for performance simulation at scale with integrated testing tools.

As AI agents increasingly collaborate with human teams, the Command Center introduces a vital observability layer, enabling users to monitor agent health, track performance, and continuously refine agents at scale. The lack of visibility into agent performance has been a significant barrier to adoption since Agentforce’s inception, particularly concerning ROI. As Ian Gotts, CEO of Elements.Cloud, previously noted, “Nobody wants to hand out a blank cheque – they want caps and predictable ROI”. With the Command Center, users finally gain a clear view of their agents’ activities, making it easier to measure impact, optimize performance, and understand their return on investment. The Command Center is slated for general availability this August.

Model Context Protocol (MCP): The Universal Connector for AI Agents

AI agents prove truly valuable when they can interact directly with a business’s existing tools to perform actions like sending files, updating records, or triggering workflows. However, securely integrating these agents without compromising security or requiring extensive custom coding has been a significant challenge. Salesforce aims to resolve this with the Model Context Protocol (MCP), described as a “universal plug, like USB-C for AI agents”. This innovative feature will enable agents to connect to diverse systems without requiring custom setups, thereby ensuring quick and secure integrations. The MCP will facilitate:

• API Conversion with MuleSoft: New MCP connectors will allow MuleSoft to transform any API or integration into an MCP-compatible service, complete with security, monitoring, and traffic control, streamlining the development and management of multi-agent workflows across an organization’s tech stack.

• Custom MCP Server Hosting with Heroku AppLink: Developers can deploy and expose custom MCP services using Heroku, offering a fast, secure, and scalable method to connect bespoke tools to Agentforce leveraging Heroku’s infrastructure and automation.

• Secure Agent Actions within Slack: Slack will soon incorporate its own MCP server, developed with Anthropic, enabling Agentforce agents to interact with Slack messages and files, providing context-rich insights and taking actions while fully adhering to enterprise permissions and data boundaries.

• Unified Agent Gateway: Agentforce introduces a governed gateway, engineered by MuleSoft, that centralizes agent registration, identity, and policy management. This allows administrators to tightly control which agents connect to specific tools, how they behave, and ensure all activities comply with enterprise-grade security and compliance standards.

The MCP is a strong indicator of Agentforce’s ambition to scale, moving beyond isolated use cases to become a fully integrated, enterprise-ready platform. This protocol ensures that agents can connect to a broader range of tools and execute real actions, all while adhering to a company’s security policies.

The Expansion of AgentExchange: A Growing Ecosystem

Building on the success of the AppExchange, Salesforce introduced the AgentExchange in March as an AI alternative, designed to help companies rapidly create and deploy AI agents for improved productivity, efficiency, and innovation across various industries. This concept perfectly aligns with the collaborative spirit of the Salesforce ecosystem. With the latest update, the AgentExchange is truly taking shape, offering high-value use cases from renowned companies. This update leverages the previously mentioned MCP servers through various partners on the AgentExchange. Salesforce has provided compelling real-world examples:

• PayPal: Service teams can now manage invoice generation via PayPal’s MCP server and agentic commerce capabilities. This enables Agentforce to process transactions, create compliant invoices, and notify customers via email with attached invoices, all without manual input.

• WRITER: By securely connecting to the WRITER MCP server, Agentforce will allow WRITER’s AI agents to automate content creation, retrieve knowledge, and perform compliance checks directly within Salesforce workflows.

• Box: Agentforce will support content teams by automatically retrieving the latest signed NDA from Box, summarizing key terms, and sending Slack updates to account teams with the document and recommended next steps.

These examples, featuring collaborations with big-name companies, represent a significant leap forward, transforming AgentExchange from an early-stage directory into a robust, scalable ecosystem.

Atlas Reasoning Engine Receives Further Enhancements

The Atlas Reasoning Engine, the core intelligence of Agentforce, is responsible for generating, evaluating, and refining plans based on an agent’s objectives. Salesforce had already updated Atlas in the 2.0 release, enhancing its reasoning capabilities with enriched data from Data Cloud, allowing it to handle more nuanced queries, loop through tools and data sources for improved answers, and utilize retrieval-augmented generation (RAG) with added metadata for increased accuracy. The 3.0 update further enhances Atlas with an “enhanced Atlas architecture”, resulting in several key improvements:

• Increased Speed and Responsiveness: Agentforce now runs up to 50% faster than it did in January 2025, with real-time response streaming allowing users to see answers as they are generated.

• Enhanced Intelligence and Accuracy: Agents can now leverage web search to find more relevant information and display inline citations for greater transparency.

• Expanded Model Choice: In addition to OpenAI, users can now utilize Anthropic’s Claude models, securely hosted on Salesforce infrastructure via Amazon Bedrock, which is particularly beneficial for highly regulated industries. Google’s Gemini will also be supported later this year.

• Greater Global and Multilingual Support: Agentforce is now available in Canada, the UK, India, Japan, and Brazil, and supports six new languages, with over 30 more on the way.

• Always-On Reliability: If one model experiences slowdowns or failures, Agentforce will automatically switch to another in the background, ensuring continuous performance.

• Government Approval: Agentforce has achieved FedRAMP High authorization, making it a trusted and available solution for U.S. public sector organizations through Government Cloud Plus.

Salesforce describes these improvements to Atlas as the “icing on the cake,” providing faster performance, smarter responses, and greater reliability across all Agentforce-built applications.

Additional Information and Final Thoughts

Agentforce 3.0 also introduces several updates aimed at helping teams quickly get started and realize value faster, regardless of their industry. These include over 100 pre-built industry actions rolling out in Summer 2025, covering diverse use cases such as patient scheduling, advertising proposal generation, and vehicle servicing. The platform now offers out-of-the-box readiness, enabling minimal setup for deploying AI agents across a wide range of sectors. Furthermore, Salesforce has simplified its pricing with new SKUs for Sales, Service, and Industry Cloud, and introduced per-user pricing with unlimited usage for employee-facing agents, making it easier to scale without worrying about usage limits.

As Salesforce continues its drive to rapidly scale Agentforce, these new features are expected to significantly help customers optimize their agents and fully harness their potential. This particular update feels like a cohesive effort by Salesforce to integrate feedback and implement impactful changes. However, the rapid pace of the “Agentforce train” has been a point of concern for many. While Agentforce 3.0 brings numerous positives, it might also be an opportune moment for Salesforce to pause, allow the dust to settle, and give users ample time to fully implement and leverage the current capabilities of Agentforce.