Salesforce and NVIDIA on September 15, 2026 announced Koa, Salesforce's first CRM-specific reasoning model — post-trained from NVIDIA Nemotron 3 Super on a proprietary synthetic dataset modeled on nearly three decades of Salesforce CRM deployments and designed to power Agentforce agents inside Salesforce's own trust boundary.
27 Years Of CRM Data, Compiled Into A Reasoning Engine
"The most valuable thing Salesforce has built isn't our platform — it's the accumulated knowledge of how enterprise business actually works," said Marc Benioff. "With Koa, that knowledge is put inside the model itself." On Salesforce's own CRM benchmark — updating opportunities, routing cases, scheduling follow-ups — Koa matches or exceeds leading model performance with three times fewer errors. No customer data was used; training scenarios were entirely synthetic, covering 14+ industries from manufacturing to healthcare.
Missionforce Brings Nemotron To Air-Gapped Government AI
Salesforce is also pulling Nemotron open models and NVIDIA accelerated computing into Missionforce, the company's federal and highly regulated product line. Missionforce Operations — used to digitize procurement, supplier management and logistics — will get post-trained NVIDIA models available to select customers in October 2026, enabling agents to run inside classified networks and fully air-gapped systems.
Pilot Customers And Availability
Koa is already running inside Salesforce (including a Slack help agent) and in pilots with 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine and Xero. General availability is expected in winter 2026 in U.S. regions. The launch is the biggest concrete piece of Salesforce's Trusted Enterprise AI Harness push and stacks on top of previously announced Accenture-Google Cloud Gemini Enterprise partnership in the agentic-enterprise race.
Reporting based on Salesforce's official press release, Unite.AI, IT Pro and Digital Today coverage.