Salesforce Introduces Agentforce 3

Salesforce introduces Agentforce 3, the next step of evoluation in how organizations deploy, manage, and scale AI-powered agents across their business. The release event showcased a unified vision for hybrid workforces, where humans and AI agents collaborate seamlessly to drive efficiency, personalization, and growth.

Here are the key takeaways from the Agentforce 3.0 release:

1. The All-New Command Centre

Agentforce 3 introduces a new Command Center that gives you a clear, real-time view of how your AI agents are performing. It’s the first tool designed specifically to monitor and manage AI agents across your business from one place.

Highlights:

• A unified pane of glass to monitor agent performance
• Track KPIs by use case (e.g., sales revenue, case resolution)
• Drill down into individual interactions to find and fix issues
• Real-time alerts and telemetry integration (e.g., with Datadog)
• Easily apply updates with natural language instructions

2. Enhanced Interoperability with Native MCP Support and Agent Exchange

Agentforce 3 makes it much easier for AI agents to work with your existing tools—securely and without custom code. With built-in support for the new Model Context Protocol (MCP), agents can now connect and interact with hundreds of external systems, tools, and data sources.

Highlights:

• PayPal, Slack, Salesforce, and more via AgentExchange
• Govern access with natural language-based guardrails
• Eliminate the need for complex custom integrations
• Admins can define what agents can do and how they interact with tools—securely and at scale.

3. Faster Time to Value: Pre-built Actions, Flexible Models, and Pricing

A core focus of Agentforce 3 is to help customers prove ROI and realise value faster. This is supported by:

• Extensive Pre-built Industry Actions: Agentforce offers over 200 curated, optimised, and industry-tailored actions. These pre-built topics and actions work out-of-the-box, built on the powerful data models available in Salesforce’s Industry Clouds, significantly reducing the effort required to get started and bridging the gap between AI’s promise and real-world impact.
• Flexible Model Choice: Salesforce is introducing its first hosted frontier model running directly on Salesforce infrastructure via Amazon Bedrock. This allows agents to reason using Salesforce-hosted Anthropic, ensuring data remains entirely within the Salesforce trust boundary – a crucial feature for customers in regulated industries.
• Flexible and Scalable Pricing: The pricing model has evolved from a per-conversation basis ($2) to a more flexible, usage-based model (10 cents per action), which is more aligned with the actual work and value agents create. Additionally, an unlimited use license at a fixed price has been introduced for employee productivity use cases, encouraging broader adoption.

References & Useful URLs

• YouTube Video (41 mts) – Salesforce Release Event