Inside Claudeforce: Uniting Salesforce Data with Claude AI

Salesforce and Anthropic deepen their strategic alliance with the launch of Claudeforce, a move that enhances enterprise connectivity and platform engagement. Credit: JasonDoiy/Getty Images

Salesforce and Anthropic expand their strategic partnership as Shannon Mathews explains how platform engagement grows through secure enterprise connections

Salesforce and Anthropic are bringing their best capabilities forward in an expanded partnership between the two, introducing Claudeforce to transform enterprise workflows.

The integration brings together Claude’s reasoning with Salesforce’s enterprise harness.

The move is aimed at making data, workflows, business logic, actions and governance securely accessible to power agentic experiences across workspaces.

We’re bringing together the world’s #1 AI and #1 CRM – the best of both worlds

Marc Benioff, Chair & CEO at Salesforce

Through this architecture, software operates as an underlying system that powers every interface.

Marc Benioff, Chair and CEO at Salesforce, says: “We’re bringing together the world’s #1 AI and #1 CRM – the best of both worlds. Today, we’re launching Claudeforce, running directly on top of Salesforce via our new AIforce UI harness, Headless 360, Data 360, Tableau and Slack.

“Here, the UI is the AI – allowing you to build custom apps dynamically and answer any enterprise question. Probabilistic intelligence alone doesn’t run a company and deterministic systems don’t reason.

“By fusing Claude’s extraordinary reasoning with the trusted data, workflows and governance every enterprise runs on, we’re delivering a dynamic interface that thinks, reasons and acts. This is how every business will run.”

Marc Benioff, Chair and CEO at Salesforce. Credit: Marc Benioff/LinkedIn

Key operational pillars

The partnership combines real-time reasoning with enterprise rules through four main integration areas:

• Salesforce inside Claude: The Salesforce in Claude plugin offers 37 pre-built sales skills that allow sellers to review pipeline details, conduct deal health reviews and prepare for meetings. Setup requires a single administrator connection with centralised authentication, giving whole teams instant access without individual account audits
  
  
• Claude inside Salesforce: Claude acts as a reasoning model for the Atlas Reasoning Engine, powering Agentforce Vibes and Agentforce Coworker by default while remaining available in Agent Builder. Through Amazon Bedrock, organisations can deploy Claude within the Salesforce Trust Boundary to maintain data security
  
  
• Slack workspace collaboration: Claude serves as the default model across Slack, powering Slackbot for individual productivity gains. It supports team decisions through Claude Tag and assists developers as a founding partner for Slack Code
  
  
• Mutual enterprise technology adoption: Both companies use each other’s technologies internally. Salesforce uses Claude as the default model across Slack AI, Headless 360, Agentforce Coworker and Claude Code for its engineering organisation. Anthropic uses Salesforce as its preferred CRM alongside Slack, Salesforce Shield, Backup and Sandboxes.

The underlying technology connecting these systems is AIforce. This enterprise harness brings business data and workflows to external agents via Model Context Protocol (MCP) servers, APIs and CLI tools without complex custom integrations.

An LLM on its own has extraordinary intelligence and reasoning capabilities. Paired with a deterministic core – like Salesforce’s trusted data, workflows and governance – it becomes something enterprises can take action on

Shannon Mathews, VP, Product Management at Salesforce

Securing external connections

As enterprise workflows shift from static screens into flexible conversational interfaces, maintaining user engagement and platform trust becomes vital.

When asked how platform engagement is protected in flexible interfaces, Shannon Mathews, VP, Product Management at Salesforce, says that the enterprise opportunity lies in making software more useful across every workspace.

She explains: “Engagement is better than ever and our opportunity is to make Salesforce more useful by making it accessible wherever people choose to work. In fact, as shared in our most recent earnings, use of the platform has surged 6x via MCP calls.”

Salesforce Earnings Report

• Annual recurring revenue from Salesforce’s core AI and data products reached nearly US$3.9bn, rising over 210% year-over-year
• Annual recurring revenue for Agentforce, Salesforce’s platform for deploying autonomous AI agents, exceeded US$1.5bn after growing more than 240% year-over-year
• Automated tasks completed by AI agents across Agentforce and Slack reached 7 billion, with 3.2 billion delivered during the second quarter alone
• Active users of Slackbot, the embedded AI assistant that automates workspace tasks, grew by over 150% quarter-over-quarter to drive record order value
• Data 360 processed 82 trillion records via Zero Copy connections, allowing AI models to analyse external data without copying it, marking a 731% surge

Shannon adds: “For decades, enterprise software required users to manually navigate static interfaces to get work done.

“In the agentic AI era, the interface becomes more flexible – while Salesforce remains the trusted system powering the data, governance, workflows, business logic and permissions underneath it.

“That’s why we recently introduced headless, which allows customers to discover and use our platform capabilities through a secure MCP connection, without rebuilding custom integrations. That’s a significant expansion of value we can deliver to our customers from the Salesforce platform.”

Shannon Mathews, VP, Product Management at Salesforce. Credit: Shannon Mathews/LinkedIn

Addressing how systems prevent hallucinations and enforce governance during autonomous actions, Shannon outlines the security architecture: “Trust is built into the architecture, not bolted on. Every answer and every action runs through your existing Salesforce permissions and business rules.

“Claude sees only what a given user is authorised to see and can only do what that user is authorised to do. You control how much autonomy Claude has. Every write back to Salesforce is grounded in rich business context and confirmed by the user.”

She notes that pairing a large language model with a deterministic core enables enterprises to execute governed actions with precision.

“An LLM on its own has extraordinary intelligence and reasoning capabilities. Paired with a deterministic core – like Salesforce’s trusted data, workflows and governance – it becomes something enterprises can take action on. Claudeforce unifies these two strengths so that Claude can reason, act and get work done with the trust and accuracy from Salesforce,” Shannon adds.

Salesforce in Claude is currently available to select pilot customers. The open beta launch is expected in September 2026, with additional pre-built skills following in late 2026.

Internal deployment of the Claude-powered Slackbot has already delivered 8.1 million hours of annualised productivity gains across Salesforce teams.

Key partners

• Salesforce: A global software provider that helps organisations become agentic enterprises by integrating humans, AI agents, applications and data on a unified platform to unlock growth and innovation.
• Anthropic: A frontier AI research organisation and public benefit corporation controlled by a Long Term Benefit Trust. The company develops frontier intelligence platforms designed to advance human progress, including its flagship Claude language models, prioritising safety, alignment and multi-step reasoning capabilities for high-stakes business environments.
• Slack: A multiplayer work operating system and productivity platform owned by Salesforce that brings people, communication, tools and autonomous agents together in a single collaborative workspace.
• Amazon Bedrock: A fully managed cloud service provided by Amazon Web Services (AWS) that offers secure API access to high-performing foundation models. Within this partnership, it hosts models like Claude inside the Salesforce Trust Boundary, allowing regulated industries to deploy domain-specific AI workloads securely without data leaving enterprise compliance controls.