Salesforce put 37 prebuilt sales skills inside Claude. Pipeline review, meeting prep and deal health now run against live CRM data, with Salesforce enforcing the rules.
If your company runs Salesforce, the question stops being which prompts to paste and starts being which skills your admin turns on. Your advantage moves from prompt craft to knowing what to ask for — and to catching output that is confidently wrong about your own deal. Reading a generated pipeline review critically will be worth more than writing a clever prompt.
A skill is a packaged method the assistant runs the same way every time, rather than a one-off prompt you retype. A plugin is the connection that lets the assistant read and write in another system, here Salesforce, under that system's existing permissions.
Salesforce and Anthropic announced Claudeforce on August 26, an expanded partnership whose first release is Salesforce in Claude — a plugin carrying 37 prebuilt sales skills, including meeting prep, deal health review and pipeline review.
The mechanics matter more than the branding. Until now, getting a model to reason about a deal meant copying context into a chat window: notes, a CRM export, a few emails. Salesforce in Claude reverses that. Claude reads live revenue context, and any action it takes routes back through Salesforce so business rules and permissions are enforced at the point of write. An admin connects it once, and every seller on the team inherits access without a new permissions model to build.
The two companies built the skills together rather than wrapping generic CRM prompts around an API. Onboarding reads a seller's Salesforce, Slack and connected sources to stand up a dashboard of their own accounts and pipeline.
The partnership runs in both directions. Claude serves as a reasoning model inside Agentforce and is the default model for Slack, Slackbot and Slack AI. Salesforce says Slackbot is driving 8.1 million hours of annualized productivity internally.
Availability is the honest caveat. Salesforce in Claude is with select pilot customers now, with open beta expected in September 2026, and additional skills arriving later in the year.
AEs: Meeting prep and deal health review stop depending on how much context you remember to paste in. Expect your manager to ask why a deal-health flag went unanswered.
SDRs: Little changes this month. The 37 skills target the revenue cycle, not top-of-funnel volume.
Sales managers: A pipeline review generated from live records changes what a forecast call argues about. Decide now whether you treat it as input or verdict.
RevOps: Actions route through Salesforce rules, so your existing validation and permission sets become the guardrail. Audit them before the beta lands.
Before the beta reaches your org, run your own version by hand. Take your three largest open opportunities, build a deal health review from the CRM record alone, and list every fact the record does not contain. That gap list is what a skill will get wrong too.
Two minutes, once a week. What changed in AI, and what to run because of it.