Salesforce opened Dreamforce in San Francisco on Tuesday by unveiling AIforce, a “live interface layer” that pipes CRM records, workflows, and permissions into outside A.I. surfaces so that sales teams can query Salesforce, log calls, and move deals forward without ever opening the Salesforce app. It’s the loudest admission yet from the category’s incumbent that the user interface it spent two decades perfecting is no longer where the work happens.
“AI is creating an interface revolution,” Marc Benioff, the company’s chair and chief executive, said at the announcement. Patrick Stokes, president of applications and marketing, put it more bluntly in a press briefing reported by IT Pro: “The UI is not actually our product.” Parker Harris, the co-founder who built the original Lightning interface, described the strategy earlier this year as “dis-intermediating” the company from itself.
AIforce launches with three surfaces. Claudeforce, a prebuilt MCP server inside Anthropic’s Claude, ships with 37 sales skills covering prospecting through pipeline hygiene, and is available to all customers in beta after pilots with Deloitte, GitLab, and Legora. Slackforce bundles Slackbot, Slackforce Surfaces, and a Slack CRM that lets a rep create an account, log call notes, and update records from a chat prompt. Agentforce Coworker, which lives inside Lightning itself, drew 100,000 users in its first 35 days.
The governance pitch is what makes the architecture legible to enterprise buyers. Every AIforce request runs on the caller’s existing Salesforce permissions, every action routes back through Salesforce, and a Zero Data Retention design blocks model providers from keeping business data. Pricing and packaging remain subject to change.
For a B2B sales team of five to thirty, the practical shift is that the CRM stops being a destination. A rep who lives in Slack can qualify an inbound lead and update its stage in the same message. That’s the friction point that historically kept small teams from adopting Salesforce cleanly at all, and it collapses only if permissions and workflows are actually encoded, which many small deployments never bothered to do.
The framing has been building for days. Salesforce quietly stripped Agentforce from its core products on the eve of Dreamforce, and last week’s public opening of the Claudeforce plug-in previewed the interface-layer thesis without naming it. Diginomica, which called the AIforce messaging “frustratingly elusive” as details trickled out, wasn’t wrong that the pitch arrived late.
But the direction is unmistakable, and it echoes a familiar move: when a platform senses its interface has become the ceiling on its own growth, it turns the interface into a protocol. Salesforce is betting that its moat was never Lightning. It was the record underneath.
Sources
- https://www.salesforce.com/news/stories/aiforce-announcement/
- https://www.itpro.com/technology/artificial-intelligence/salesforce-wants-enterprises-to-break-free-of-a-shared-user-interface-with-aiforce
- https://www.techradar.com/pro/live/dreamforce-2026-live-were-in-san-francisco-for-salesforces-big-ai-event
- https://diginomica.com/dreamforce-2026-may-aiforce-be-you-whatever-it
- https://www.unite.ai/salesforce-unveils-aiforce-live-interface-layer-at-dreamforce/
