TL;DR — Key Takeaways
- Salesforce unveiled AIforce as a new interface layer that lets users access Salesforce data, update records and run workflows from external AI platforms without opening Salesforce directly.
- New integrations include Claudeforce for Anthropic’s Claude, Slackforce, Google Gemini Enterprise, NVIDIA’s Nemotron technology and Siemens.
- AIforce is designed to preserve existing Salesforce security permissions while operating under a Zero Data Retention framework.
SAN FRANCISCO – Salesforce Inc. kicked off its annual Dreamforce conference here on Tuesday with a major architectural shift, unveiling a new interface layer designed to decouple its core business data from traditional browser-based layouts and embed it directly into third-party artificial intelligence (AI) tools.
AIforce serves as a live connective layer, allowing enterprise users to query Salesforce data, update records, and execute complex workflows from within external platforms such as Anthropic’s Claude, Slack, and Agentforce Coworker — all without opening a standard Salesforce dashboard.
The strategy arrives as CEO Marc Benioff seeks to position the enterprise software giant at the center of a rapidly evolving agentic landscape.
“AI is creating an interface revolution,” Benioff said during his opening keynote. “We are combining model intelligence with all the context that customers have built into Salesforce to create an intelligent, dynamic, composable system that is securely governed.”
AIforce serves as the latest component in Salesforce’s broader architecture, joining its Data 360 platform and Customer 360 suite.
According to the company, every request made through AIforce adheres to a customer’s existing security permissions and operates under a Zero Data Retention framework, ensuring vendor models do not store sensitive corporate data.
Salesforce debuted a handful of key integrations at the conference.
Claudeforce, developed in partnership with Anthropic, embeds Salesforce intelligence natively into Claude via a prebuilt Model Context Protocol (MCP) server. It ships with 37 prebuilt sales skills alongside developer-focused plug-ins.
Slackforce enables Slack users to generate dynamic data views, log call notes, and deploy an upgraded Slackbot capable of automated cross-platform account management.
Integrations were also announced with Google’s Gemini Enterprise, NVIDIA Corp.’s Nemotron open-source technology (powering a new model named Koa), and Siemens.
“Benioff made three moves at Dreamforce that point in the same direction. Salesforce opened its data, logic, and permissions to AI work surfaces like Claude and agents running in interfaces it does not own. It post-trained an open-weight NVIDIA base into a CRM reasoning model and kept the result inside its own trust boundary. And it put finished agents on the price list next to the tooling for building them,” said Mitch Ashley, vice president and practice lead for Software Lifecycle Engineering and AI-Native Software Engineering at The Futurum Group.
“That is a move across every layer where value could settle, suggesting Salesforce is making multiple bets rather than aligning on a single approach,” Ashley said. “Owning the connector layer and selling completed work are the two strong positions in that set. In the coming weeks, we’ll see the capabilities of their Koa model and how it benefits customers.”
“Salesforce is preparing for a world where people may work in Claude, Slack, or another AI interface without opening Salesforce at all,” said Stephanie Walter, practice leader for AI Stack & Enterprise Application Development at HyperFRAME Research. “AIforce is designed to let those interfaces use Salesforce’s customer data, business logic, permissions and actions. Salesforce is trying to remain the system behind the work even when it no longer owns the screen.”
“AIforce raises an important question: who will own the interface where employees actually work? Salesforce may remain essential underneath, but the company that owns the screen can shape which tools people use, which workflows get built and where new software spending goes,” Walter said.
The announcement coincides with the 20th anniversary of Salesforce’s Trailblazer developer community, which now counts 12 million custom apps built on its platform.
The product launches precede an investor day on Wednesday, where analysts expect management to detail long-term financial targets, clarify metrics for its new Agentforce Apps and Data 360 segment, and update its pricing strategy as the company transitions from a traditional software-as-a-service model toward consumption-based pricing for AI agents.
While Salesforce shares surged nearly 23% following a strong second-quarter earnings report in late August, the stock dipped over 1% in premarket trading Tuesday to $254.95, reflecting broader market caution. Analysts note that while Salesforce’s AI tools are gaining momentum, core business segments like Tableau and MuleSoft face lingering pressure amid shifting enterprise IT budgets.
The technology push comes amid growing scrutiny of the rapid pace of AI deployment. Activists have organized a “Stop The AI Race” protest for Thursday, the conference’s final day.
Demonstrators plan to rally outside the headquarters of OpenAI and Anthropic before marching to San Francisco City Hall to demand local leaders declare an “AI Emergency.” The movement underscores rising public and political anxiety surrounding AI risks, even as tech leaders gather inside to champion the technology’s future.
