Claudeforce makes the Headless 360 strategy concrete.

Claudeforce is the expanded partnership between Salesforce and Anthropic. Salesforce in Claude is the first product under that partnership: a plugin with 37 sales skills for live revenue context, pipeline updates, meeting preparation, deal review, and governed action from inside Claude.

Treat the announcement as a strong architecture signal. Treat pilot access and planned beta dates exactly as Salesforce describes them.

Claude supplies frontier reasoning. Salesforce in Claude packages that reasoning for sellers. Actions route through Salesforce so the company’s existing permissions and business rules continue to govern what each person and agent can see and do.

Salesforce says the product is enabled by AIforce, its enterprise harness for exposing business data and workflows through MCP servers, APIs, and CLI tools. Headless 360 makes Salesforce capabilities available outside the traditional interface, while Data 360 provides trusted customer context.

That separation matters. The model and work surface can change without moving permissions, validation rules, customer history, or process controls into a collection of one-off integrations and prompts.

The interface becomes Claude. Salesforce remains the governed system that holds business truth and executes the action.

• The frontier model can change. The business capabilities and governance underneath it remain reusable.
• Salesforce can serve as the action and trust layer for agents working in Claude, Slack, Agentforce, ChatGPT, or a custom experience.
• A user’s existing identity and permissions become part of the agent architecture instead of an integration detail handled later.
• People can keep judgment, approval, and exception work while agents carry the routine updates and analysis between decisions.

Salesforce’s August 2026 agentic AI study reinforces the same operating lesson. Clean and accessible data, tightly bounded agent scope, and human escalation paths established in advance were the strongest predictors of successful deployments. Product access cannot replace that preparation.

1. Choose one revenue workflow where faster context or cleaner follow-through creates measurable value.
2. Audit the Salesforce objects, permissions, sharing rules, validation, and automations that the agent would inherit.
3. Define the trusted context the workflow needs. Fix the relevant customer and pipeline data instead of waiting for every company record to be perfect.
4. Set write authority, approval points, human escalation, and session review before an agent acts against production data.
5. Design for a portfolio of agents and work surfaces. Claude may be one interface while Slack, Agentforce, and custom agents handle other parts of the work.

A prebuilt connection cannot clean an existing permission model, repair untrusted data, or decide which parts of a role should change. That remains the transformation work.

Dreamforce 2026 runs September 15 through 17. These are the announcements heading into the event, not a recap of announcements made on the Dreamforce stage.

• Claudeforce product pageProduct description, sales skills, FAQ, and current availability↗
• Salesforce and Anthropic announcementPartnership structure, AIforce, Salesforce in Claude, and availability↗
• Expanded Headless 360 capabilitiesMCP, Data 360, Headless Experience Layer, Skills, and Slackbot status↗
• Slack Code announcementShared code channels, supported agents, oversight, and availability↗
• Salesforce agentic AI ROI studyCurrent research on data readiness, scope, escalation, adoption, and ROI↗