Dreamforce was quite a doozy last week, and if you weren't there, you missed a monumental shift in the Salesforce ecosystem.
The central theme wasn't just AI, but the Agentic Enterprise. We saw Salesforce shift AI from a simple tool, to AI as an active, autonomous teammate, and how the entire platform is shifting for building and managing these AI agents. The key takeaway? AI is no longer a feature, it's the foundation.
Before you dive into the full article with my deep-dive notes on the Agentic Enterprise, I want to invite you to a special live session on Salesforce Security sponsored by EZProtect. The session will be answering the question on How can IT Security teams build Salesforce domain expertise without burning out?
When: This Thursday, 10/23 at 12PM ET/ 9AM PT. You can RSVP by clicking the image below:
TechieZach’s Dreamforce Recap
The rebrand of core clouds to Agentforce XYZ
Salesforce renamed its core clouds to reflect the new focus. I’m not going to list all the rebrands here, that is why websites such as RenameForce exist. Using sales cloud as an example, this has now been rebranded to Agentforce sales. You can repeat this with Agentforce Marketing, and Agentforce Service. Oddly enough, Data Cloud was not rebranded to Agentforce Data, but instead was rebranded to Data 360.
Agentforce Vibes now available!
Agentforce Vibes was a big feature throughout Dreamforce. The tool is available in any VS Code-compatible IDE, including Cursor or Mulesoft’s Anypoint Code Builder.
At its core is Vibe Codey, an autonomous AI agent that acts as a pair programmer. It understands your specific Salesforce org's schema, metadata, and existing code base, allowing it to generate, refactor, and adhere to your company's existing coding standards.
It handles the full development lifecycle from prompt to deployment. Developers can use natural language to plan features, generate Apex and Lightning Web Components, create required unit tests (to meet the 75% coverage minimum), debug issues, and manage deployments via tools like DevOps Center.
Unlike consumer-grade coding assistants, Agentforce Vibes is backed by the Salesforce Trust Layer, ensuring built-in security, compliance, and governance controls. This makes the AI-generated code auditable and safe for building production-ready business applications.
Slack as the Agentic OS
Salesforce significantly elevated the role of Slack, positioning it as the Agentic Operating System (OS) for the enterprise, serving as the central hub for human and AI agent collaboration.
Slack is being positioned at the Agentic Operating System for the enterprise, with the capability to layer all your agents into one platform.
Real-Time Work: Employees can command AI agents, pull reports, view dashboards, and access CRM data directly within Slack conversations without switching platforms.
Instant Action: Data 360 insights can automatically surface in a Slack thread, giving teams data-driven suggestions and the ability to trigger actions in other enterprise systems.
OpenAI Integrations in Slack: The expanded partnership with OpenAI included a ChatGPT app for Slack and the availability of @codex for developers to generate, review, and debug code directly in the collaboration platform.
Launch of Mulesoft Agent Fabric
MuleSoft Agent Fabric is an agent management platform designed to solve the challenge of "agent sprawl" in the enterprise, where numerous AI agents are built across different teams and vendors, leading to fragmentation, compliance risks, and operational chaos. The tool is announced to be GA October 2025, sometime following Dreamforce.
A new tool to govern your agents across your enterprise
The key features of Mulesoft Agent Fabric include:
Agent Registry: A centralized catalog for discovering and reusing every AI agent and tool across the enterprise. It serves as a unified location, like an Anypoint Exchange for your digital workforce, to prevent duplication and ensure visibility.
Agent Governance: Provides enterprise-grade guardrails and security by applying compliance and policy controls to every agent-to-agent and agent-to-system interaction, often through the Anypoint Flex Gateway. This is crucial for enforcing policies like PII detection, rate limiting, and access control.
Agent Broker: The intelligent routing service that orchestrates complex workflows. It uses Large Language Models (LLMs) for reasoning and planning, dynamically discovering the right specialized agents from the registry and ensuring they work together to complete multi-step tasks seamlessly and securely.
Agent Visualizer: A real-time monitoring dashboard that provides end-to-end visibility and observability into the agent network. It gives IT teams a dynamic map of how agents are interacting, helps trace their decision-making process, and surfaces key performance metrics (like confidence scores) for optimization and auditing.
Data Cloud Keynote Recap
Data 360 is positioned as the critical, trusted data layer that grounds all of the new AI agents within the Agentforce 360 platform, ensuring they have the necessary context to act safely and accurately. Improvements across the board were noted for structured and unstructured data alike.
Zero Copy Technology: Data 360 offers a unified data foundation that can instantly access data from hyperscalers like Snowflake and Google BigQuery without physically duplicating the data. This provides a single, real-time view while improving data governance and reducing costs.
Context Engine for Agents: Data 360 acts as the "context engine" that feeds both structured and unstructured data in real time to the Agentforce 360 AI agents. This context is what allows the agents to reason, predict, and take intelligent, personalized actions.
Foundation for Agentforce Voice: Data 360 provides the grounding for the new Agentforce Voice feature, ensuring that voice-based AI agents operate with full data context, execute actions like updating CRM records, and maintain complete auditability for every action taken.
P.S Don’t forget to join me this Thursday (10/23) for a live discussion with Matt Meyers on How can IT Security teams build Salesforce domain expertise without burning out? RSVP link.