Jensen Huang at Dreamforce: AI Will Help Us Know Everything and Do Anything
NVIDIA founder and CEO Jensen Huang brought an expansive vision for artificial intelligence to Salesforce’s Dreamforce conference Tuesday, joining Salesforce CEO Marc Benioff onstage to discuss the rapid evolution of AI, the future of enterprise software, and the importance of building safe and reliable systems.
Huang’s appearance coincided with Salesforce’s announcement of Koa, the company’s first CRM reasoning model. Built on NVIDIA Nemotron 3 Super, Koa is designed to support complex customer relationship management tasks and bring more advanced reasoning capabilities to Salesforce’s Agentforce platform.
The Jensen conversation between Huang and Benioff was energetic and informal. At one point, Huang stepped away from the stage and walked through the audience at San Francisco’s Moscone Center with a microphone in hand while Benioff followed alongside him.
“We’re all following you, bro,” Benioff told Huang, prompting applause from the audience.
But the Jensen central message of Huang’s appearance went beyond the latest product announcement. He described AI as the next major infrastructure layer in the evolution of technology, comparing its potential impact with earlier industrial and technological revolutions.
“With electricity, we could power everything,” Huang said. “With the internet, you can find anything. And now, with artificial intelligence as an infrastructure layer across the planet, we can know everything and do anything.”
That vision provided the broader context for Koa, which represents Salesforce’s effort to develop specialized AI reasoning capabilities for enterprise CRM applications.
Building a Reasoning Model for Enterprise CRM
Koa was developed by post-training NVIDIA Nemotron 3 Super using a proprietary synthetic dataset based on nearly three decades of enterprise CRM deployments. The approach allows Salesforce to tailor the model specifically to the types of tasks businesses perform across sales, customer service, and other CRM workflows.
Because NVIDIA Nemotron is an open model, Salesforce was able to fine-tune and operate it within its own infrastructure. The company maintains control over the model weights, while the model runs on Salesforce systems without using customer data during training or inference.
“Not a single byte of customer data was used,” said Rohan Kumar, Salesforce’s president of platform and engineering, while introducing Koa ahead of Huang’s appearance.
The Jensen company used supervised fine-tuning and reinforcement learning as part of the development process, relying on NVIDIA’s NeMo RL, NeMo Gym, and NeMo AutoModel technologies.
Koa’s training corpus includes synthetic enterprise scenarios spanning more than 14 industries, including manufacturing, financial services, healthcare, and travel. The goal is to give the model a stronger understanding of the specialized workflows and actions required in business environments.
Salesforce says Koa has already demonstrated strong performance on its CRM Bench, a benchmark designed around real-world CRM activities such as updating opportunities, routing customer cases, and scheduling follow-ups.
According to Salesforce, Koa matches or exceeds leading models on CRM actions while producing three times fewer errors.
AI Safety as an Engineering Priority
AI safety was another major focus of the conversation. When Benioff asked Huang about the issue, the NVIDIA CEO emphasized that safety must remain central to AI development and deployment.
“Safety is paramount in a lot of ways. It’s job one,” Huang said.
At the Jensen same time, he described AI safety as an engineering challenge rather than an issue that should prevent innovation.
“However, safety is an engineering problem,” Huang said. “We’re developing computing systems after all.”
Huang argued that companies should not release products or services when they lack confidence in their functionality, capabilities, or safety. He also rejected the idea that organizations must choose between moving quickly and building safe technology.
“Innovation, speed and safe products — it’s a false choice,” he said. “You could definitely have both at the same time.”
His message was that companies should continue moving rapidly while maintaining the discipline to stop and address problems when necessary.
“So run as fast as you can,” Huang said. “But if you feel at any given point in time the company’s out of control, or the product’s not going to be safe, take a pause and make sure you get it right.”
The Rise of the AI Enterprise
Huang also used the Dreamforce stage to discuss his broader expectations for how AI will transform businesses and the software industry.
He Jensen pushed back against concerns that AI would eliminate the software industry, noting that he had previously challenged that assumption. He also rejected the idea that AI should simply be viewed as a threat to employment.
Instead, Huang emphasized the potential productivity gains that businesses could achieve by adopting AI.
“As a result of our ambition, and with the productivity boost we get from AI, the sky’s the limit,” he said. “Engage AI. Don’t get left behind.”
Huang Jensen predicted that AI adoption will eventually extend across companies, enterprises, and countries.
“Every company, every enterprise, every country would become an AI company,” he said. “It’s going to be an agentic enterprise.”
The concept of the agentic enterprise centers on AI systems capable of taking actions and completing tasks rather than simply generating information in response to prompts.
Benioff echoed that vision while highlighting Salesforce’s customer and developer community.
“The sky is especially the limit for our Trailblazers,” Benioff said. “I think this technology, the way it empowers people … the technology can partner with them to do this in incredible new ways.”
Open Models Gain Momentum
Huang also highlighted the growing adoption of open AI models. According to him, the share of AI models being built or adopted using open approaches has increased significantly.
“Open models went from 30% at the beginning of last year to now some 70%,” Huang said.
He noted that organizations are pursuing multiple AI strategies at the same time. While businesses continue adopting closed models, enterprises and software companies are increasingly developing customized AI systems designed around their own data, workflows, and requirements.
“People are both adopting closed models at exponential rates, but also building their own custom AIs — because every single software company is an AI company, and every single enterprise is going to be an agentic and AI enterprise,” Huang said.
Jensen NVIDIA itself is using Salesforce across sales, service, marketing, and operations. The company is also piloting Agentforce for customer-support workflows, giving NVIDIA firsthand experience with the technology it is helping enable.
Koa Moves Toward Customer Deployment
Koa is already being used within Salesforce, including to power an employee agent in Slack. The company plans to expand access through customer pilots beginning in October, when Koa will become a customer-selectable model within Agentforce.
The Jensen initial customer group includes Formula 1, UChicago Medicine, Baxter Credit Union, 1-800Accountant, Engine, and Xero.
Salesforce expects Koa to reach general availability in U.S. regions in winter 2026.
The Jensen announcement represents another step in Salesforce’s broader push to integrate reasoning and autonomous capabilities into enterprise software. By combining a model tailored specifically for CRM tasks with Salesforce’s existing Agentforce platform, the company aims to move AI beyond conversational assistance toward systems capable of carrying out meaningful business workflows.
For Jensen Huang, however, the significance of developments such as Koa extends beyond CRM. He sees AI becoming a foundational layer of the global technology infrastructure, much like electricity and the internet before it.
His Jensen Dreamforce message was ultimately centered on that transition: AI is moving from an emerging technology into an essential component of how organizations operate, create software, serve customers, and make decisions. The companies that embrace that transformation while maintaining strong engineering and safety standards, Huang suggested, will be positioned to explore an increasingly broad range of possibilities.
Source Link: https://blogs.nvidia.com/