Salesforce spent two decades training a generation of sellers to live inside its screens. This week it said the screens are now optional.
Salesforce and Anthropic announced Claudeforce on August 26, an expanded partnership that puts Claude’s reasoning directly on top of Salesforce’s data, workflows and business logic. The flagship product, Salesforce in Claude, ships with 37 prebuilt sales skills, meeting prep, deal health review and pipeline review among them, that let a seller run their entire revenue cycle from inside a Claude conversation instead of a CRM screen. It is live for select pilot customers now, with an open beta planned for September.
The pitch: software becomes a system, not a screen
The framing Salesforce is using is deliberately blunt. “For decades, enterprise software required users to manually navigate static UI to get work done. Now that agents can access the data, workflows, and rules directly, software is a system that powers every interface,” the company said in its announcement. Marc Benioff, Salesforce’s chair and CEO, put it more directly: “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.”
Anthropic’s framing is about trust rather than interface design. “We believe frontier intelligence should be safe, trusted, and deeply capable, which is why the world’s leading businesses turn to Claude for their most important work,” said Dario Amodei, Anthropic’s CEO and co-founder. “Salesforce in Claude brings this same frontier intelligence into the systems where much of the world’s commercial activity happens. Through this partnership, companies can point Claude at the customer information and business context that they’ve been building in Salesforce for decades, and use it to actually run and grow their businesses.”
What actually ships
The mechanics matter more than the messaging. Salesforce in Claude is built on AIforce, described as Salesforce’s enterprise harness that exposes business data and workflows to any agent through MCP servers, APIs and CLI tools, rather than requiring a custom integration for each one. An admin connects it once, and every seller on the team gets access from day one, with authentication and permissions managed centrally instead of rebuilt per user.
On first connection, the plugin reads a seller’s context across Salesforce, Slack and any other connected source, then generates a dashboard of that seller’s accounts and live pipeline inside the conversation itself. Actions taken through Claude still route through Salesforce’s own business logic, so a discount approval or a stage change still has to clear the same governance rules it would if a human had clicked through the UI. That routing choice is the part worth watching: it means Salesforce is not ceding control of what an agent can do, only how a seller gets to that action.
Why this is bigger than one CRM
Salesforce is not alone in wiring its product into a frontier model’s own interface rather than the reverse. Smaller platforms have been making the same bet at a lower level of ambition: Wix recently put an AI agent in charge of routine SMB marketing tasks inside its own product. The direction of travel across enterprise software this year has been toward agents that sit on top of a company’s systems and act with governed authority, not just chat about them. What makes Claudeforce notable is which company is doing the ceding: Salesforce built its business on being the system of record sellers had to open every morning. Handing that entry point to a third-party model, even one it now depends on commercially, is a bet that owning the data and workflow layer matters more than owning the screen.
It is also a hedge against a narrative Salesforce has been fighting publicly. Benioff has pushed back on what he called the “SaaSpocalypse” story, the idea that general-purpose AI models make specialized business software obsolete. Claudeforce is Salesforce’s answer: if a frontier model is going to be where work happens anyway, better to be the trusted data and governance layer underneath it than to be replaced by a thinner one.
What it means for the marketing leader
Sales is the first workflow Claudeforce touches, with marketing, service and commerce capabilities coming in the months after. For a CMO, the near-term signal is less about the specific skills shipping today and more about the pattern: the systems marketing already depends on, campaign platforms, CDPs, attribution tools, are all candidates for the same move. If a rep can review a deal by asking Claude instead of opening an opportunity record, a campaign manager will eventually be able to ask a model to build a segment, launch a journey or diagnose a drop in attribution without opening the platform’s own dashboard.
That has two practical implications worth planning for now. First, vendor evaluations should start asking not just what a platform’s UI does, but what its API and agent-access layer expose, since that is increasingly where the actual work will happen. Second, governance built for humans clicking through screens, approval chains, audit trails, permission tiers, needs to be re-checked against a world where an agent is the one initiating the action. Salesforce’s answer was to route every agent action back through its existing business rules rather than let the model act unchecked, the same governance question ad tech has been wrestling with as agents start transacting on their own. Any martech platform following the same path should be able to show the same thing.
What to watch next
The open beta in September is the first real test of whether “the UI is the AI” holds up outside a curated pilot. Watch whether the 37 sales skills expand cleanly into marketing use cases, and whether other CRM and martech vendors respond with their own agent-access layers rather than betting their own chat interface can compete with Claude’s. The vendors that lose the most ground here are unlikely to be the ones building agent integrations. They are more likely to be the ones still asking customers to open a dashboard to find out what already happened.
Source: Salesforce
