Salesforce AIforce Explained: How CRM Data Moves Into Claude, Slack and Any AI Interface

Salesforce AIforce is a live interface layer announced at Dreamforce on September 15, 2026. Its purpose is to expose the data, workflows, business logic, semantics, permissions and governance already inside Salesforce to AI interfaces outside the traditional CRM screen.

In practical terms, AIforce is Salesforce’s attempt to make CRM capabilities follow the user. Instead of opening a Salesforce dashboard, a worker can ask a question in Claude or Slack, retrieve governed CRM context, update records or trigger workflows from the interface they are already using. Salesforce says AIforce launches with Claudeforce, Slackforce and Agentforce Coworker.

The key distinction is architectural: Agentforce provides the agents; AIforce provides the interface layer that brings Salesforce context and actions to those agents and other AI tools.

AIforce is less about adding another chatbot and more about changing where enterprise software lives. If Salesforce’s model works as described, the CRM stops being only a destination and becomes a governed back-end for AI interfaces that employees already use.

The names are close enough to be confusing. Salesforce describes Agentforce as its digital workforce: agents that can reason, plan and execute tasks across Salesforce and connected systems. AIforce sits above and around that layer by making Salesforce capabilities available to multiple AI interfaces.

This distinction matters for buyers because adopting AIforce does not eliminate the need to design, test and govern Agentforce agents. It changes how people and external AI systems can reach the capabilities Salesforce already exposes.

Claudeforce: Salesforce inside Claude

Salesforce in Claude is the clearest example of the AIforce strategy. Salesforce says the integration uses a prebuilt MCP server so Claude can work with Salesforce context and actions without every customer building the connection from scratch. The initial package includes 37 prebuilt sales skills, while a Salesforce Development plug-in for Claude Code offers more than 40 development skills.

Salesforce says Salesforce in Claude has been piloted by companies including Deloitte, GitLab and Legora and is now available to all customers in beta. Future plans include Tableau analytics plus skills for service, marketing, commerce and industry workflows.

Slackforce: CRM context inside team conversations

Slackforce is positioned as the collaborative surface. Salesforce says Slackforce Surfaces can turn live context from Salesforce, Slack and other tools into interactive interfaces that teams can filter, explore and act on in real time. Slackbot can reason across Slack conversations plus Salesforce context and governed actions.

A practical example is an account-retention workflow: Slackbot could identify accounts that are going quiet, inspect support cases and relevant conversations, reassign ownership, create a follow-up task and draft outreach without making the employee jump through several CRM screens.

Agentforce Coworker: AI inside Lightning

Coworker is the Salesforce-native side of the launch. It runs inside Lightning, can call specialized Agentforce agents already deployed in the organization, and operates under existing permissions and business rules. Salesforce says Coworker is available to customers and can be activated without creating a new permission model.

Underneath the product branding is a broader architectural direction Salesforce has been building throughout 2026. In April, the company introduced Headless 360, describing Salesforce capabilities as accessible through APIs, MCP tools and CLI commands rather than only through browser interfaces. AIforce turns that headless approach into a user-facing enterprise AI layer.

Salesforce says AIforce is powered by the Headless Toolkit, exposing platform capabilities through MCPs, APIs, plug-ins, skills and developer tools. Salesforce’s architecture documentation also describes MCP as a bridge for connecting agents to external systems, while Agent2Agent (A2A) communication supports interactions between agents.

That creates a useful mental model:

1. The model understands the request.
2. AIforce provides access to governed Salesforce context.
3. Agentforce or other tools perform permitted actions.
4. Salesforce permissions and business rules remain in the execution path.
5. The result is returned in the user’s chosen interface.

The promise is portability of interface without throwing away the CRM’s existing controls. Whether that works cleanly in a specific deployment depends on how permissions, skills, integrations and external tools are configured.

Salesforce says AIforce requests run through existing permissions and business rules, so an AI interface should only see what the requesting user is allowed to see. The company also describes the launch architecture as using Zero Data Retention with model providers for business data used to answer a request.

Those are important controls, but they do not remove customer responsibility. Salesforce’s own Agentforce security guidance uses a shared responsibility model: Salesforce provides foundational security controls, while administrators remain responsible for access configuration, agent guardrails and the permissions granted to tools and users.

For enterprises, four checks matter before expanding AIforce broadly:

• Permission inheritance: verify that AI interfaces respect the same record-, object- and field-level boundaries expected in Salesforce.
• Action scope: separate “read and summarize” permissions from actions that update records, send messages or trigger workflows.
• External tool trust: understand which model provider, MCP server or agent receives which data during each step.
• Auditability: ensure administrators can reconstruct who requested an action, which agent or skill performed it and what changed.

In other words, AIforce can reduce interface friction while increasing the number of paths into sensitive enterprise systems. That makes governance more important, not less.

Salesforce’s announcement is specific about several launch components but deliberately cautious about commercial terms. Salesforce in Claude is available to all customers in beta, and Salesforce says Agentforce Coworker is available to Salesforce customers. Slackforce features are part of the broader launch and rollout, but exact availability can vary by product, region and customer agreement.

Salesforce did not publish a single universal AIforce list price in the launch announcement. The company states that pricing and packaging are subject to change and that availability may vary by region. That means buyers should avoid treating AIforce as one SKU with one simple price.

Salesforce has previously introduced flexible Agentforce pricing using Flex Credits, but that historical pricing model should not be assumed to represent the full cost of an AIforce deployment without a current quote.

1. Sales teams that live outside the CRM

A salesperson who spends most of the day in Slack or Claude can query pipeline context, inspect account history, log notes or trigger follow-up workflows without opening a separate Salesforce screen. That can reduce context switching, although organizations still need disciplined data quality.

2. Service teams combining conversations and CRM history

Support work often spans tickets, account records, knowledge bases and internal discussions. An AI interface that can reason over governed context from those systems could reduce time spent assembling the case history before taking action.

3. Managers building temporary interfaces

AIforce’s “composable interface” idea is especially interesting for one-off operational questions. Instead of waiting for a fixed dashboard, a manager may be able to describe the view needed and receive an interactive interface built from governed data.

4. Developers building agentic workflows

For builders, the Headless Toolkit and MCP approach can make Salesforce capabilities available inside coding agents and external AI tools. The benefit is not merely developer convenience; it is a standardized path for bringing business rules and permissions into those tools.

AIforce solves a genuine usability problem, but it can create new operational failure modes if companies equate easier access with safe automation.

• Bad CRM data becomes easier to act on. AI cannot fix incomplete ownership, duplicate records or outdated pipeline stages by itself.
• Prompt ambiguity can become an action problem. “Move this forward” may sound clear in conversation but can be unsafe if the agent can update multiple records or trigger a workflow.
• More interfaces mean more governance surfaces. Claude, Slack, Coworker and future integrations all need consistent policies.
• Beta features change. Organizations should not design mission-critical processes around roadmap assumptions.
• Human review still matters. High-impact actions such as changing pricing, approving refunds or sending regulated communications need deterministic checks.

The safest rollout is incremental: start with read-only or low-risk workflows, validate permissions and logs, then expand action scope as the system proves reliable.

AIforce is part of a wider move from software as a destination to software as a service layer for agents. Microsoft, Google, OpenAI, Anthropic and major SaaS vendors are all trying to let AI systems retrieve enterprise context and take actions across existing tools.

Digital Pulse Brief has already covered a related shift in ChatGPT Work’s Data Agent for company analytics and the growing importance of agent monitoring and model-misalignment controls. AIforce approaches the same problem from the CRM side: keep business data, permissions and workflows in Salesforce while allowing multiple AI interfaces to use them.

The strategic question for enterprises is therefore not “Which chatbot should we buy?” It is “Which system should remain authoritative when many AI interfaces can act on our data?” Salesforce’s answer is that CRM context, permissions and execution should remain governed by Salesforce even when the user interface moves elsewhere.

Salesforce-heavy organizations with mature permissions and clean CRM processes are the most obvious early candidates. They already have the data, workflows and governance AIforce is designed to expose.

Teams already using Claude or Slack may see faster value because the first integrations are targeted directly at those environments.

Organizations with weak access controls or inconsistent CRM data should fix those foundations first. AIforce can make good workflows easier to access, but it can also make bad data and overly broad permissions easier to propagate.

No. AIforce is an interface and platform layer. Salesforce separately announced Koa, its first CRM reasoning model built on NVIDIA Nemotron. AIforce can connect AI interfaces to Salesforce context and actions regardless of which supported model or agent sits on top.

No. Agentforce is Salesforce’s platform for building and deploying agents. AIforce brings the governed Salesforce platform—including Agentforce capabilities—to interfaces such as Claude, Slack and Coworker.

Salesforce says Salesforce in Claude is available to all customers in beta as of the September 2026 announcement. Availability can still depend on customer agreements and region.

Yes. Salesforce says the Headless Toolkit behind AIforce exposes platform capabilities through MCPs, APIs, plug-ins, skills and developer tools. Salesforce in Claude uses a prebuilt MCP server.

Salesforce did not announce one universal AIforce list price. The company says pricing and packaging are subject to change and availability varies. Customers should obtain a current quote based on the specific Salesforce, Slack, Agentforce, Data 360 and partner components they need.

• Salesforce: AIforce announcement — September 15, 2026.
• Salesforce and Anthropic: Claudeforce announcement.
• Salesforce: Headless 360.
• Salesforce Architects: Agentic patterns and implementation.
• Salesforce Help: Agentforce security and shared responsibility.
• Salesforce and AWS enterprise AI expansion.

• ChatGPT Work Gets a Data Agent: How OpenAI Wants Teams to Turn Company Data Into Answers
• OpenAI Model Misalignment Framework: 6 Incidents and What They Mean for AI Safety
• Explore more Business Technology coverage