AI agents are becoming more capable, but most remain isolated from the systems, data sources, and workflows organizations require for day-to-day operations. For regulated enterprises operating at massive scale, agents must do more than generate answers. They must run securely, integrate with governed data, control cost, and execute actions inside business workflows.<\/p>\n\n\n\n
Salesforce and NVIDIA are addressing this challenge for enterprises by bringing high-performance, cost-efficient AI agents into the flow of work. Powered by Agentforce, Slack, and NVIDIA Nemotron<\/a> models \u2014 part of the NVIDIA Agent Toolkit<\/a> \u2014 organizations can now deploy governed agents across regulated and on-premises environments while enabling employees to use them directly in the flow of their work.<\/p>\n\n\n\n At the core is Agentforce, Salesforce\u2019s platform for building and orchestrating enterprise AI agents. Agentforce grounds agents in trusted data through Data 360, applies business logic, and coordinates actions across Customer 360 applications and connected systems, ensuring governance, compliance, and predictable performance at scale.<\/p>\n\n\n\n NVIDIA Nemotron 3 Nano is available within Agentforce, providing a model optimized for enterprise scale and efficiency. Its 1 million token context window<\/a> enables agents to reason across long customer histories, large documents, and complex multi-step workflows without losing context. Built on a Mixture of Experts architecture<\/a>, it increases computational efficiency, reduces reasoning tokens, and lowers compute demand and cost in multi-step agent workflows.<\/p>\n\n\n\n For regulated industries, customers can deploy Nemotron models on NVIDIA AI infrastructure in on-premises or tightly controlled, private-cloud environments, allowing organizations to keep model processing within their own security boundary while meeting strict compliance and data residency requirements.<\/p>\n\n\n\n Slack connects these agents to the people doing the work. Slackbot acts as the coordination layer for those agents. It receives requests from users in Slack, triggers Agentforce workflows, and orchestrates agent actions across enterprise systems. A request in Slack can trigger an Agentforce workflow, reason over Data 360 context, invoke Nemotron powered processing, and execute approved retrieval actions across enterprise systems before returning recommendations into the conversation.<\/p>\n\n\n\n For a regulated financial services company, this could mean a compliance agent reviewing transactions, applying policy rules, and surfacing risk signals in a secure Slack channel. For a healthcare organization, it could mean summarizing extensive case histories at scale while enforcing data access controls defined within Salesforce.<\/p>\n\n\n\n Enterprise AI adoption often slows down due to uncertainty about how orchestration, data grounding, infrastructure, and collaboration fit together.<\/p>\n\n\n\n Salesforce and NVIDIA are developing reference architectures that clearly define each layer:<\/p>\n\n\n\n These patterns give companies a practical blueprint. They show how prompts originate in Slack or Customer 360 applications, how Agentforce coordinates reasoning and actions, how Data 360 provides context, where model processing runs, and how responses return into business workflows. This removes ambiguity and enables customers to implement secure, scalable agent solutions with confidence.<\/p>\n\n\n\nNVIDIA Nemotron Models Available in Agentforce<\/strong><\/h2>\n\n\n\n
Clear Architecture for Companies<\/strong><\/h2>\n\n\n\n
\n