ARTHUR GOLDSTUCK | Salesforce brings in Claude to power the enterprise

Claude is fast becoming the AI model business wants to work with. Every time I give a talk on AI in the enterprise and ask for an AI choice show of hands, Anthropic’s model moves further ahead of OpenAI’s ChatGPT.

No surprise, then, that customer relations software giant Salesforce this week announced it would meet Claude in that same space.

The result of the collaboration between the two leaders is the delightfully named Claudeforce. Unusually, it is a corporate mash-up word that works. A platform rather than a product, its first offering is Salesforce in Claude, a plug-in with 37 prebuilt sales skills. A seller can ask Claude to prepare for a meeting or assess the health of a deal using live Salesforce data. Claude can also take actions back into Salesforce, subject to the permissions and business rules already in place there.

In effect, Salesforce controls what happens once Claude tries to act on an answer.

“Here, the UI [user interface] is the AI — allowing you to build custom apps dynamically and answer any enterprise question,” says Marc Benioff, chair and CEO of Salesforce. “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.”

The irony is that Salesforce has built one of the world’s biggest enterprise software businesses by getting companies to work inside Salesforce. Claudeforce is a signal that Salesforce accepts that users may begin their work inside Claude.

In practice, that means a salesperson could ask which accounts need urgent follow-up or why a deal has slowed. Claude can pull the relevant Salesforce information into the response and, where authorised, update the underlying record. The employee may barely see Salesforce, yet rely on it throughout the task.

Salesforce calls the architecture behind this AIforce. It makes business data and workflows available to AI agents through application programming interfaces (APIs) — a set of rules that lets different software programs talk to each other — and Model Context Protocol (MCP), a common standard for connecting AI systems to external tools and information.

This is where Benioff’s distinction between probabilistic and deterministic systems kicks in. Claude can interpret an untidy request and work through the context, while Salesforce enforces the company’s permissions and workflows when the request turns into action.

The Anthropic tie-in is well timed, with the AI company preparing for what is likely to be a record-breaking IPO, thanks to becoming one of the strongest contenders for enterprise use of generative AI.