A new CRM reasoning model to power your digital workforce
Introducing Koa, Salesforce’s first CRM reasoning model for Agentforce, built on NVIDIA Nemotron™. Koa is trained with 27 years of Salesforce CRM intelligence to enable agents to reason through the complex, multi-step tasks required for enterprise work.
Trained on 27 years of Salesforce workflows
Purpose-built models
Koa is grounded in nearly three decades of CRM deployments, built by post-training NVIDIA Nemotron open models with a proprietary synthetic dataset modeled on how businesses actually run: their processes, workflows, and operational policies. Developed through deep technical collaboration with NVIDIA, it's purpose-built to deliver deeper expertise on targeted enterprise tasks.
Fine-tuned for complex CRM work
Koa’s training corpus is built entirely from synthetic scenarios simulating the reasoning, tool use, and decision-making Agentforce agents perform across the customer lifecycle — from generating leads to resolving service cases — spanning 14+ industries like manufacturing, financial services, healthcare, and travel. Each scenario paired a persona with specific tasks and mapped the exact sequence of actions needed to complete them.
Kept inside the trust boundary
NVIDIA Nemotron's open models gave Salesforce the control to build CRM intelligence directly into the model itself. Salesforce controls the weights and runs Koa entirely within its own infrastructure — so no customer data ever crosses the trust boundary during training or inference. The result: a specialized, Salesforce-hosted model purpose-built to power your Agentforce use cases.
Salesforce is customer zero, using Koa to test and benchmark across a variety of CRM use cases, such as help agents, employee agents, event agents, web agents, and more. Koa's performance is measured on Salesforce's CRM Bench, a benchmark built from real-world tasks like updating an opportunity, routing a case, and scheduling a follow-up.
11% more precise at calling the right action so agents take fewer wrong turns, leading to less customer frustration
Recalls customer context with 2.1 times greater reliability delivering higher relevance and faster answers
15% better at remembering context in long back-and-forth conversations, so customers don’t have to repeat themselves
Select Salesforce Koa as best fits your organization. It’s available as a managed LLM in your generative AI models catalogue, selectable org- wide in Agentforce as a model provider, and of course as a model in Agentforce builder at the agent and sub-agent level.
Any agent, sub-agent, or agent router can use Salesforce Koa – giving you the maximum flexibility in how you leverage Koa.
Koa becomes the fourth model provider options in Setup, applied across all agents in an org. Customers opt-in, just like any other provider.
Koa joins Data Cloud Generative Models catalogue, available to all customers out of the box. Usable in AI apps, prompts, and agents immediately.
Koa is already in use inside Salesforce, including our Employee Agent that helps employees find information and complete everyday tasks in Slack, and is now moving into customer pilots with 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine, and Xero.
“Some of the most important work in healthcare happens behind the scenes, coordinating information, navigating complex processes, and making sure the right next step happens at the right time. The [Placeholder Name] reasoning model can work across those longer, multi-step workflows and help our teams manage that complexity more effectively. That creates more time and capacity for what matters most, caring for patients.” – Andrew Chang, Chief Marketing Officer, UChicago Medicine
“Our members come to us with goals, whether that’s buying a home, managing their money, or planning for what’s next. Koa can help our Digital agents understand the full complexity and context behind those goals and reason across the information, tools, and policies needed to move them forward. It gives us a powerful way to make every interaction more intelligent, personal, and useful.” – John Sahagian, SVP and Chief Data Officer, Baxter Credit Union
“At Engine, we're building the future of business travel, and that starts with equipping our teams to deliver exceptional service. Business travel has countless moving parts, and what we need from AI isn't a model that sounds confident — it's one that can reason precisely through complex, multi-step problems. Collaborating with Salesforce and NVIDIA on a reasoning model purpose-built for Agentforce lets us push that vision further, faster.” – Elia Wallen, Founder and CEO, Engine
“Accounting requires navigating tax rules, financial data, documents, and the unique circumstances of every customer. Koa gives our agents the reasoning to work through that complexity step by step and use the right tools along the way. That means we can extend more of our accountants’ expertise across every customer interaction and help people get to the right outcome faster.” – Ryan Teeples, Chief Strategy Officer, 1-800Accountant
Salesforce Koa is Salesforce’s first CRM reasoning model for Agentforce, built on NVIDIA Nemotron™. You can select it as a managed LLM in the Data Cloud generative models catalogue, enable it org-wide in Agentforce as a model provider in Setup, or apply it as a standard model in prompt builder at the Agent and Sub-Agent level.
Salesforce Koa is being used internally to test and benchmark across a variety of CRM use cases – surfacing across help agents, employee agents, event agents, web agents, and more. In Salesforce’s CRM Bench, a model benchmark that includes a suite of real-world tasks like updating an opportunity, routing a case, or scheduling a follow-up, Koa already matches or exceeds leading model performance on CRM actions with 3x fewer errors. Compared to today’s default general intelligence models we are seeing: A) 11% more precise at calling the right action so agents take fewer wrong turns, leading to less customer frustration, B) 2.1 times greater reliability delivering higher relevance and faster answers, and C) 15% better at remembering context in long back-and-forth conversations, so customers don't have to repeat themselves.
Salesforce Koa is a new AI model option customers can use in Agentforce. It can be selected for use at the org level, in Agentforce setup, and for each agent and sub-agent.
Salesforce Koa only ever sees what you choose to share, so you stay in full control of what context is provided. As with other managed models at Salesforce, Salesforce Koa draws on the grounding data, records, and instructions you make available.
No customer data is ever used in the training of Salesforce Koa. Its training corpus was built entirely from synthetic scenarios that reflect the reasoning, tool use, and decision-making skills Agentforce agents perform across CRM workflows: generating leads, qualifying opportunities, and resolving service cases across the customer lifecycle.
Salesforce Koa has enterprise-grade security at multiple levels that govern how the model behaves and where it runs. First, Koa itself lives entirely inside Salesforce's trust boundary, so your data always stays under your ownership and control. Second, Koa is optimized to follow your instructions and stay within the guardrails you set rather than improvise beyond them. We host the model at ‘temperature 0’ so responses remain consistent and repeatable instead of open ended. Third, we have a dedicated ‘serving harness’ which adds trust and safety controls, so your data is protected by more than the model alone. The result is Salesforce Koa – an enterprise-grade reasoning model that keeps you in control and delivers Salesforce's trusted infrastructure at every level.
Salesforce Koa is available to select pilot customers now and expects to be generally available (GA) in U.S. regions in Winter 2026 – with plans for an open beta starting shortly after.