Salesforce Announced Koa Reasoning Model for CRM
The new AI model uses synthetic data to improve multi-step workflow accuracy for CRM agents.
Updated on Sept. 23, 2026 in Artificial Intelligence
Live Poll
Do you trust AI-driven automation to handle complex tasks in your professional dealings?
Salesforce has unveiled Koa, a specialized reasoning model designed to assist agents with complex multi-step workflows. Built on the NVIDIA Nemotron 3 Super model, Koa aims to enhance CRM performance while operating within strict security boundaries.
Why it matters
Koa addresses the need for reliable AI agents capable of executing multi-step CRM actions without exposing sensitive user information. By using synthetic datasets rather than customer data, the model provides a secure alternative for enterprise automation.
Koa leverages synthetic scenarios across 14 distinct industries to simulate tool use and complex workflows. The model incorporates Supervised Fine-Tuning and reinforcement learning with Group Relative Policy Optimization.
The players
Salesforce
This company is a global leader in customer relationship management software and enterprise cloud computing solutions.
NVIDIA
This technology corporation specializes in the design and manufacture of graphics processing units and artificial intelligence hardware.
The details
The model is currently deployed for specific Slack tasks and has undergone pilot testing with organizations including 1-800Accountant, BCU, Engine, Formula 1, UChicago Medicine, and Xero. Salesforce also confirmed that Missionforce Operations for government agencies will feature integrated NVIDIA Nemotron models.
Timeline
September 16, 2026: Salesforce and NVIDIA officially announced the Koa model.
October 2026: Post-trained NVIDIA models become available for select Missionforce customers.
Winter 2026: Koa is expected to reach general availability in the United States.
The Tech Race
The introduction of Koa reflects a broader industry shift toward specialized, domain-specific AI models that prioritize reasoning over generic generative text. By moving away from training on live customer data, Salesforce is positioning itself to compete with foundational model providers who face increasing data privacy scrutiny.
Enterprise users can expect a reduction in errors when executing multi-step CRM tasks as the model integrates into their daily workflows. The focus on trust boundaries means that businesses can utilize these AI-driven reasoning tools without compromising their internal data privacy.
The takeaway
Koa highlights the increasing importance of synthetic data training to bypass the privacy risks associated with large-scale CRM datasets. Readers should monitor how these reasoning models impact the precision of automated service tasks as they roll out to wider enterprise markets.
What happens next
General availability for the Koa model is scheduled for Winter 2026, with select Missionforce customers receiving access to post-trained NVIDIA models in October 2026.
Further reading
For more information on the evolution of enterprise automation, visit the Artificial Intelligence section.
Live Poll
Do you trust AI-driven automation to handle complex tasks in your professional dealings?
