Salesforce and Nvidia have announced the launch of Koa, a new AI reasoning model built specifically for completing customer relationship management tasks within the Agentforce platform.
Salesforce said Koa was built by post-training Nvidia’s open-weight AI model Nemotron 3 Super using synthetic data modelled to replicate CRM deployments.
It is designed to provide assistance to humans within the Salesforce platform as well as power agents to automate tasks such as generating leads, closing sales, and resolving support tickets.
In an internal benchmark Salesforce made to test models against their capacity to complete agentic tasks within a CRM environment, Koa achieves similar scores to OpenAI’s GPT-5.6 Terra and Anthropic’s Claude Opus 4.6, at a faster speed and lower cost.
Salesforce said the model produces three times fewer CRM errors than competing models and is trained and inferenced entirely within the existing Salesforce trust boundary so sensitive data is not shared.
Marc Benioff, chair and chief executive of Salesforce, said: “The most valuable thing Salesforce has built isn’t our platform – it’s the accumulated knowledge of how enterprise business actually works.
“With Koa, the knowledge is put inside the model itself. We trained a reasoning engine that understands the structure of a deal, the lifecycle of a service case, and the workflows that vary across industries. That’s a different kind of intelligence, and it runs entirely inside your trust boundary.”
Salesforce added that it has already deployed the model within its cloud-based communication platform Slack, to power an agent that helps its own employees to complete daily tasks and search for enterprise information.
It has also made Koa accessible within Missionforce, its specialised platform for defence, intelligence, and national security organisations. This allows the model to complete tasks in controlled environments such as private cloud and air-gapped networks.
.jpg "Rogue agents breached Indian organisations says country’s cyber authority")