Salesforce launches AIforce to push CRM logic into Claude, Slack and every AI interface

Salesforce unveiled AIforce on September 15, 2026, at its Dreamforce conference in San Francisco, describing it as a live interface layer that exposes the company's data, workflows, business logic, semantics, permissions and governance to any external AI interface. The product launched with three initial surfaces: Claudeforce, built with Anthropic; Slackforce, which lives inside Slack; and Agentforce Coworker, an AI teammate inside Salesforce's own Lightning interface. Salesforce chair and chief executive Marc Benioff framed the announcement as an interface revolution rather than a product refresh.

The launch arrived against a turbulent backdrop for enterprise software. SiliconANGLE reported on September 15 that Salesforce stock had been under pressure for much of 2026 as investors worried that AI agents would erode the value of traditional applications, a narrative that became known as the SaaSpocalypse. Benioff opened the keynote by rejecting that idea outright, calling it crazy nonsense in remarks covered by The Next Web, and arguing that the shift was not about the end of software but about the end of software that makes humans do all the work.

Salesforce is not positioning AIforce as a replacement for its core platform. Unite.AI reported on September 15 that the layer removes the need to open the Salesforce application at all in order to reach business knowledge: a worker who has never opened a dashboard can ask questions, update records and trigger workflows grounded in Salesforce data and business logic, while agents can read across hundreds of records at once. Users can compose custom live interfaces on the fly by describing them inside Slack, Claude or the Coworker surface.

The groundwork for this push was laid earlier in 2026. CIO reported on September 15 that Salesforce introduced Headless 360 at its TDX developer conference in March, an API driven layer for agent first workflows that exposed MCP servers and APIs. Patrick Stokes, Salesforce president of applications and marketing, described Headless 360 as the raw ingredients and AIforce as the chocolate chip cookie that comes out of the oven. By the company's own account, Headless 360 proved difficult to use and scale, which created the opening AIforce is meant to fill.

Key Facts

AIforce ships with Claudeforce, Slackforce and Agentforce Coworker. SiliconANGLE reported on September 15 that Claudeforce extends the Salesforce partnership with Anthropic and brings 37 prebuilt sales skills covering everything from prospecting to pipeline hygiene, with Tableau analytics plus service, marketing, commerce and industry skills coming soon. A plug-in for Claude Code adds 40 skills, access to a Salesforce skills library on GitHub, and dynamically loaded sub plug-ins.

Slackforce lives inside Slack and consists of Slackforce Surfaces, which are live interactive interfaces, Slackbot, an out of the box assistant, and Slack CRM, which lets users open accounts, log call notes and update records without leaving Slack. Agentforce Coworker is an AI teammate inside Salesforce that calls the specialized Agentforce agents a business has already deployed. The Next Web reported on September 15 that 100,000 users turned on Agentforce Coworker within its first 35 days.

Salesforce also used the keynote to launch Koa, its first CRM reasoning model. The Next Web reported on September 15 that Koa was post trained on Nvidia's Nemotron 3 Super using a synthetic dataset drawn from 27 years of CRM deployments, with not a single byte of customer data used, according to Salesforce president of platform and engineering Rohan Kumar. Salesforce says Koa matches or beats leading models with three times fewer errors on its internal CRM benchmark, controls the weights and runs inference on its own infrastructure. Pilot customers include 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine and Xero.

Several large numbers surfaced around the launch. Salesforce said internally it runs 7,000 users on Claudeforce, that its Hunter sales agent generated 500 million dollars in pipeline last quarter, and that its Casey service agent has resolved 5 million conversations. Fulton Bank senior vice president Brandon Bernard said the bank went from zero to more than 20 use cases live in production within weeks, supporting approximately 3,000 users. Elia Wallen, founder and chief executive of Engine, said his company handles over 800,000 customer inquiries a year.

Analysis

The strategic logic behind AIforce is straightforward once you separate the interface from the system of record. Salesforce is conceding that the place where work happens is moving away from its own screens, and it is trying to make sure that whichever AI surface wins, the data, permissions and business rules underneath still come from Salesforce. What this really means is that the company would rather be the governed substrate beneath Claude, Slack and the next agent interface than defend a login page that knowledge workers may never need again.

That is a real bet, and it carries real risk. The more value that migrates into third party interfaces, the harder it becomes for Salesforce to control the customer relationship, the pricing conversation and the upsell motion. Benioff's answer is architectural rather than contractual: AIforce runs on the customer's existing permissions and business rules, with no new permissions model, no migration and no custom integration work, per Unite.AI. Under the zero data retention design championed by Benioff, business data answers the question at hand and is not retained by the model provider. If that governance story holds up under scrutiny, it becomes the moat that keeps customers anchored even as the interface layer dissolves.

The bigger picture here is that Salesforce is testing whether an enterprise software company can survive the disappearance of its own user interface. Cofounder Parker Harris, who designed the Lightning interface, has questioned whether users would need to log into Salesforce at all, and Stokes told CIO that the UI is not Salesforce's product, that the product is the platform where customers store their data, metadata, workflows and permissions. That is a striking thing for a company whose revenue has long been tied to seats and screens to say out loud.

It also explains the speed of the Anthropic and Nvidia alignments. Claudeforce was announced with Anthropic in August and reached open beta on the AppExchange as part of this launch. The Next Web reported on September 15 that Salesforce shares were up more than 50 percent from a 52 week low of 146 dollars in late June, and had gained more than 4.5 percent on the Monday before the keynote after Anthropic chief executive Dario Amodei called for AI development to slow down. The market is clearly trading on the Anthropic relationship as much as on Salesforce's own roadmap.

Why It Matters

If AIforce works, it changes what enterprise software is for. Instead of asking humans to learn a labyrinth of menus, the platform becomes a governed service that answers questions, updates records and triggers workflows on demand. Unite.AI reported on September 15 that AIforce becomes the newest layer in Salesforce's Agentic Enterprise architecture alongside Data 360 for unified context, Customer 360 for semantic intelligence and Agentforce for a deployable digital workforce. Every request runs on existing permissions and business rules, which is the part that enterprise buyers will actually test.

The reach matters too. CIO reported on September 15 that Salesforce wants to broaden its business by reaching knowledge workers inside large organizations who have never used Salesforce before, letting them create custom, composable applications that treat Salesforce as the system of record. That is a new addressable market for a company that has spent decades selling to sales, service and marketing teams. Analyst Daniel Newman, chief executive of The Futurum Group, said core ERP, CRM and HR tools will see a similar realignment, which suggests Salesforce is early to a wave rather than alone in it.

There is also a community dimension. Unite.AI reported on September 15 that the rollout coincides with the 20th anniversary of the Trailblazer community, which Salesforce said has built 12 million apps, drives 9.6 billion API calls per day, has produced 156 million lines of AI generated code and has earned more than 151 million badges on Trailhead. If composable interfaces become the norm, that ecosystem of builders is the distribution channel for the next generation of Salesforce work.

Next Up

Salesforce has promised more. Benioff said an SDK for building AIforce apps is coming, along with an AIforce Max Edition that bundles everything together, while Stokes said additional AI platforms would be announced soon. CIO reported on September 15 that the Claudeforce open beta is already available on the AppExchange, and Stokes said customers can grab the plug-in in open beta, install it into a Claude enterprise account and have the MCP servers, skills, security and zero data retention wired up almost instantaneously. He also warned that pricing, packaging and regional availability may change.

The open questions are the ones Salesforce cannot answer from a keynote stage. Will large customers trust a third party model provider with governed access to their CRM, even with zero data retention? Will developers build on AIforce instead of going directly to the underlying APIs? And will the seat based economics that built Salesforce survive a world where the interface is described in a sentence inside Slack or Claude? The answers will arrive over the next several quarters, one product cycle at a time.