Welcome to the New Era of Agentforce & Salesforce AI

Welcome! Whether you are an administrator, developer, or platform architect, keeping pace with Salesforce's rapid pivot toward agentic execution can feel like moving at lightspeed. We have moved far beyond basic chat assistants into an ecosystem where autonomous agents execute multi-step workflows, reason dynamically over unified data, and operate securely side-by-side with your teams.

This hub serves as your centralized briefing to track what’s new, understand core platform shifts, and stay ahead of the latest capabilities across Agentforce and Salesforce AI.

Summary of Latest Changes & Platform Updates

Recent platform updates focus heavily on tightening developer control, streamlining multi-agent architectures, and expanding real-time voice and automation capabilities:

• Agent Optimizer (Formerly Agentforce Assistant): The built-in AI helper within Agentforce Builder has been renamed to Agent Optimizer (with the interface shifting to Ask Optimizer), providing a clearer UX for building, troubleshooting, and tuning agent instructions.
• Centralized Conversational Memory Control: Admins can now globally toggle and manage conversational memory across all agents and users via a single org-wide setting, streamlining privacy and state governance.
• Reusable Skills for Agentforce Coworker: You can now package specific instructions, context, and tool invocations into discrete, reusable Skills built via Skill Builder, making it simple to scale multi-step capabilities across agents.
• Expanded LLM Ecosystem: General Availability (GA) rollouts include high-performance production models like Claude Haiku 5.5, Claude Sonnet 5.5, and GPT-6.1 Sol (each at a 1,000 requests/minute rate limit), giving orgs greater flexibility in prompt templates and backend reasoning.
• Real-Time Voice Streaming & Auto-Disconnect: Amazon Connect can now stream live call audio directly to Agentforce over WebSockets without requiring static phone numbers or complex SIP trunks. Plus, new Auto-Hangup and Auto Disconnect parameters allow agents to cleanly conclude calls and pass intent-fulfillment metrics back to telephony partners.