Salesforce Partners with OpenAI for Government AI

Salesforce on September 16, 2026 announced new Missionforce capabilities that give government agencies greater control over how they develop, customize, and deploy AI, alongside a new Missionforce partnership with OpenAI to bring OpenAI’s AI models into secure government environments.

Salesforce said the new capabilities are designed to reduce manual work for government teams and support always-on missions while streamlining policy execution, enhancing operations, and improving citizen support. It said the partnerships are intended to unlock immediate operational impact: decision clarity for operators in the field, greater efficiency and mission readiness for agencies, and faster, more responsive services for citizens.

“Government agencies want tailored AI that runs everywhere they operate while staying within the secure environments their missions demand,” said Kendall Collins, CEO of Missionforce & Government Cloud. Collins said Missionforce enables this at scale, with purpose-built AI that executes the mission and lets agencies solve their hardest problems while maintaining operational control.

OpenAI Integration Through Government Cloud

Under the partnership, OpenAI’s frontier models will integrate with Salesforce’s Public Sector Solutions through Amazon Bedrock via Salesforce Government Cloud, the company said. Missionforce applications and workflows will also be available through ChatGPT, allowing government and national security personnel to access secure, authorized data and trigger Missionforce actions from a familiar chat interface. Salesforce said these capabilities together give agencies the ability to reason across mission data and execute multistep workflows.

Salesforce described the partner strategy as pairing Missionforce capabilities with U.S.-origin models and trusted AI and mission infrastructure, and said government agencies will gain frontier-level AI performance while maintaining operational control. Alongside the OpenAI partnership, the company said it is bringing NVIDIA models and accelerated computing into Missionforce so customers in highly sensitive industries can train, tune, and deploy mission-specific models on their own critical data. Salesforce and NVIDIA are also working to bring post-trained NVIDIA models into additional Missionforce capabilities, enabling customers to build and run models on their own infrastructure for specialized workloads.

Policy Engine, Operations, and Field Operations

Salesforce said the new Missionforce Policy Engine uses OpenAI’s generative AI capabilities to transform approved policy documents into structured, deterministic rules. It is built on an open-source rules engine with human-readable, open-source rule definitions, turning static regulations into mission-ready workflows so agencies can keep policy execution aligned with regulatory changes while delivering faster, more consistent services to the public.

OpenAI’s models will integrate with the Policy Engine to generate deterministic rule code from approved policy documents, along with test cases designed to validate the rules’ accuracy and expected behavior. Every output will remain subject to human review and approval before deployment. The Policy Engine will also integrate with ChatGPT, giving citizens a way to find answers to policy questions and helping government employees retrieve accurate, up-to-date guidance when supporting constituents.

As a stated example, Salesforce said that when a family’s income changes, a state Medicaid program can automatically reevaluate a child’s coverage eligibility under current rules, with the system determining coverage within minutes and generating a full audit trail showing how the decision was made.

According to Salesforce, Missionforce Operations turns manual, paper-bound processes, including procurement, supplier management, and invoice audits, into digital workflows in minutes. Specialized AI agents automatically orchestrate tasks, track status, and flag exceptions in real time, running entirely within private cloud or air-gapped environments. The company will also fine-tune the NVIDIA models that power agents for Missionforce Operations. Trained on an organization’s operational data and terminology, the specialized models are designed to let agents reason through back-office processes and act within the customer’s environment, including air-gapped networks.

In Salesforce’s stated example, AI agents can scan multi-depot inventory systems within an air-gapped environment during a critical fleet maintenance surge, identify matching spare parts, and generate transfer manifests and priority dispatch schedules to help reduce asset downtime from days to minutes.

Salesforce said Missionforce Field Operations & Asset Management lets government agencies use AI agents to automate scheduling, work orders, and asset maintenance, supported by native offline mobile capabilities for remote environments. In the company’s stated example, following a natural disaster the agents can automatically schedule follow-up work orders so local site inspectors and emergency response teams can conduct safety assessments and expedite repairs to restore power, clean water, and public infrastructure.

Earlier NVIDIA Announcement and Availability

The expansion follows the September 15, 2026 announcement of Koa, which the release describes as Salesforce’s first CRM reasoning model for Agentforce, built on NVIDIA Nemotron. That release detailed the NVIDIA collaboration inside Missionforce, extending control over model, data, and deployment environment to government and regulated organizations across private clouds, air-gapped networks, and other secure environments. According to the release, Missionforce Operations is generally available in U.S. regions, with post-trained NVIDIA models available to select customers in October 2026.