Salesforce pushes agentic AI deeper into government, with OpenAI in the mix

Salesforce is sharpening its pitch to government buyers, announcing new Missionforce AI capabilities and a partnership with OpenAI that it says will bring frontier models into secure government environments.

The announcement was made by Salesforce on 16 September 2026. It is primary material from the company, not independent validation.

Missionforce is Salesforce’s agentic platform for government, pitched as a bundle of AI, data, apps and cloud technology for public sector and defence work. Salesforce said the new capabilities are designed to give agencies more control over how they develop, customise and deploy AI, while streamlining policy execution, operations and citizen support.

The OpenAI partnership is the headline grabber. Salesforce said OpenAI’s models will integrate with Public Sector Solutions through Amazon Bedrock via Salesforce Government Cloud. It also said Missionforce applications and workflows will be available through ChatGPT, giving authorised government and national security personnel a chat interface for accessing secure data and triggering Missionforce actions.

Salesforce also pointed to work with NVIDIA. According to the company, it is bringing NVIDIA models and accelerated computing into Missionforce so customers in sensitive industries can train, tune and deploy mission-specific models on their own critical data.

A key new feature is Missionforce Policy Engine. Salesforce said it uses OpenAI’s generative AI capabilities to turn approved policy documents into structured rules and test cases. The company said outputs will still require human review and approval before deployment.

That detail matters. Government AI is not just another chatbot race. Agencies need audit trails, permissions, security boundaries and clear accountability. Salesforce’s own framing leans heavily on those points, with repeated references to operational control, secure environments and human approval.

In a separate Salesforce post, also dated 16 September 2026, the company marked one year of Missionforce and claimed public sector had become its fastest-growing industry business at scale. It said its defence sector business grew 80 per cent year on year.

Salesforce also said it now supports government work across more than 30 countries, 60 US federal agencies, all 15 US Cabinet departments, all 50 US states and all six branches of the US military.

The company listed several US government deployments. It said the US Army awarded Salesforce a 10-year indefinite delivery, indefinite quantity contract with a US$5.6 billion ceiling to expand Missionforce National Security. It also said Army Human Resources Command selected Missionforce National Security for AI-powered HR support covering 9.2 million soldiers, veterans, civilian staff and military families.

Salesforce claimed the US Air Force 441st Vehicle Support Chain Operations Squadron deployed Missionforce National Security for a US$13.5 billion fleet of more than 84,000 vehicles. It also said the Transportation Security Administration’s Ace AI agent handles about 100,000 traveller conversations a month and resolves 96 per cent of routine inquiries without human escalation.

Salesforce is also using commercial examples to sell the same broader agentic AI story. In a 27 July 2026 customer story, the company said Live Nation built an Agentforce-powered venue support agent across 120 venue websites, with projected automation of more than 300,000 fan inquiries annually.

Taken together, the message is blunt: Salesforce wants to be the platform where AI agents move from demos into operational systems, including high-stakes government work. The claims are Salesforce’s, but the direction of travel is clear.