Context and the road to Agentforce 3
Salesforce’s Agentforce platform has evolved at break‑neck speed since its debut at Dreamforce in September 2024. In just nine months the company shipped Agentforce 1, 2.0, and 2dx, each step moving from a handy AI co‑pilot toward a fully‑fledged agentic platform. The latest release—Agentforce 3 (23 June 2025)—directly tackles the two biggest blockers early adopters reported: visibility (knowing what agents are doing) and control (governance, security & observability).
What’s new in Agentforce 3
| Feature / Advancement | Highlights |
|---|---|
| Agentforce Command Center | Real‑time observability hub: monitor latency, error & escalation rates, receive live alerts, replay entire sessions, and analyse adoption & cost metrics. Low‑code testing tools let admins generate agents and simulate load before go‑live. |
| Model Context Protocol (MCP) | Native support for the open standard sometimes dubbed “USB‑C for AI agents”. MuleSoft connectors turn any API into an MCP server, Heroku AppLink hosts custom servers, and Slack will expose its own MCP server— all governed through a new unified agent gateway. |
| Expanded AgentExchange | 100+ new pre‑built industry actions plus a catalogue of MCP servers from 30+ partners (AWS, Box, Cisco, Google Cloud, IBM, Notion, PayPal, Stripe, WRITER, etc.). Plug‑and‑play services mean agents can process payments, retrieve NDAs, or generate compliant content without custom code. |
| Enhanced Atlas Reasoning Engine | 50 % lower latency, streaming responses, retrieval‑augmented answers with inline citations, automatic model fail‑over, and choice of hosted Anthropic Claude today (Google Gemini later in 2025). Now authorised at FedRAMP High for U.S. Public Sector. |
| Global availability | New regions: Canada, U.K., India, Japan, Brazil. Adds French, Italian, German, Spanish, Japanese & Portuguese (with 30+ more languages coming). |
| Flexible pricing & actions | New per‑user SKUs for Sales, Service, Industry Cloud with unlimited employee actions. 200+ industry actions (100+ brand‑new) accelerate time‑to‑value across healthcare, advertising, automotive, and more. |
| Cost transparency | Digital Wallet integration shows exactly how many credits each agent consumes so leaders can pause or optimise usage. |
| Agentforce Data Library | One‑click drop‑down in Setup connects agents to structured or unstructured data—no full Data Cloud build‑out required. |
Influencer perspectives
Marc Benioff – CEO & Co‑founder, Salesforce
Benioff has become the platform’s chief evangelist, repeatedly describing Agentforce as the next frontier of digital labor. He told Bloomberg that “AI is doing 30–50 % of the work at Salesforce now” and predicted a $3‑12 trillion market for digital labor. While proud of the platform’s reported 93 % accuracy, he concedes that closing the remaining gap will take sustained effort. Benioff frames Salesforce as “Customer Zero”, running the world’s largest agentic deployment on itself to prove the concept.
Adam Evans – EVP & GM, Salesforce AI
Evans summarises the release simply: “We’ve unified agents, data, apps and metadata to create a digital labor platform … Agentforce 3 brings greater intelligence, performance and trust.” After thousands of deployments, the new tools focus on observability, control and interoperability—the pain points customers raised most often.
Sanjna Parulekar – SVP, Product Marketing
Parulekar calls the Command Center “full observability and governance down to the individual interaction”. She positions MCP within Salesforce’s “open but coherent” philosophy—connect anything, but keep integration meaningful and secure. To lower the learning curve, Salesforce introduced the Agentforce Data Library and published best‑practice guides.
Customer & analyst voices
- Ryan Teeples, CTO, 1‑800Accountant – reports 70 % of peak‑season chats resolved autonomously, crediting the new observability layer for confident scale‑up.
- Mollie Bodensteiner, SVP Operations, Engine – says native MCP support lets her team connect agents to core systems “without custom code or compromising governance.”
- Athina Kanioura, Chief Strategy & Transformation Officer, PepsiCo – highlights unified insights through Data Cloud and faster strategic action.
- Martin Schneider, Analyst – notes the shift from single agents to multi‑agent orchestration and warns buyers to demand deeper analytics before scaling.
Why it matters
Agentforce 3 pushes autonomous agents from hype to enterprise‑grade reality by adding the missing ingredients—observability, governance and plug‑and‑play integrations. Faster responses, global reach and flexible pricing reduce friction; the Digital Wallet and Command Center finally expose ROI. Yet leaders should adopt with eyes open: accuracy is high but not perfect, and success still hinges on data quality, measurement and human oversight.