AI-generated editorial illustration; not a documentary photograph.
Salesforce announced two additions to its enterprise AI portfolio at Dreamforce on September 15, 2026: Koa, a reasoning model designed for customer relationship management workflows, and Claudeforce, an Anthropic-powered offering within the new AIforce interface layer.
The announcements extend Salesforce's partnerships with Nvidia and Anthropic as the company builds more autonomous functions into Agentforce. The report did not specify pricing, rollout timing or measured performance for either product.
Koa Targets CRM Workflows
Koa was built with Nvidia using the Nemotron open-weights platform. Salesforce says the model can help Agentforce agents reason through complex, multistep CRM workflows. Its training used a large synthetic dataset derived from nearly 30 years of Salesforce's internal experience with CRM deployments, according to company executives.
Salesforce said it deliberately used synthetic data instead of direct customer information during training. Chief Executive Marc Benioff also said customer data submitted while using Koa will not be used to train the model further. The source provides no technical evaluation of those privacy controls or Koa's reasoning capabilities.
Claude Gains Prebuilt Skills
Claudeforce brings Anthropic's Claude model into AIforce with 37 prebuilt skills for sales-management tasks, including prospecting and maintaining pipeline data. It follows earlier work that integrated Claude with Agentforce 360 and introduced AI features for Slack users.
Benioff said Salesforce intends to work with multiple model providers, including Anthropic and OpenAI. He described the Anthropic relationship as partly shaped by industry circumstances, noting that Salesforce had previously sought to invest in OpenAI but could not do so after Microsoft's major investment.
AI Safety Returns to the Stage
Separate Dreamforce discussions focused on responsibility and oversight. Anthropic Chief Executive Dario Amodei called for better industry practices and common standards, while Benioff argued that accountability should apply across AI companies rather than imposing a uniform slowdown on the entire sector.
OpenAI Chief Executive Sam Altman acknowledged concerns about how advanced AI systems could be misused or escape intended controls, while expressing confidence in OpenAI's ability to develop them safely. These remarks were statements from company leaders, not evidence of new safeguards or regulatory commitments.
SIGNAL TITANS
Leave a Reply