Salesforce just signed a contract worth up to $1.6 billion with the US Department of Veterans Affairs

TL;DR

Salesforce signed a contract worth up to $1.6 billion with the VA, funding agents that answer calls and schedule appointments. The agents also route cases, verify benefits, and support community care scheduling, with no diagnosis or treatment decisions. The $1.6 billion is a ceiling: one base year plus two optional one-year renewals, decided annually. My conclusion: agentic AI in healthcare scales first in operating work like intake, routing, verification, and scheduling.

Salesforce just signed a contract worth up to $1.6 billion with the US Department of Veterans Affairs. The agents it funds will answer calls and schedule appointments.

The full scope: surfacing knowledge during live calls, routing and triaging cases, automating benefits verification, supporting community care scheduling. No diagnosis, no treatment decisions. Unglamorous work, and precisely where agentic AI in healthcare scales first.

The architecture behind the contract

The architecture backs that up. VA is buying Agentforce Health and Public Sector together with Slack, MuleSoft, Data 360, and Tableau. Agents deployed on top of an integrated data and workflow layer. That combination is the product.

The deployment base already exists. VA Health Connect has handled more than 40 million calls. Slack runs across more than 150 VA medical and outpatient centers. The relationship goes back more than a decade.

And register which company this is. Salesforce spent two decades at the edge of life sciences, letting Veeva own pharma CRM on top of its platform. With that partnership ending, the entry is now coming from both sides: Viatris and Pierre Fabre signed onto Agentforce Life Sciences this spring, and the VA agreement plants the same agent stack inside care delivery.

The fine print on the numbers

The fine print: $1.6 billion is a ceiling. One base year, two optional one-year renewals. The VA gets an annual decision on whether the promised operational gains are materializing.

The boldest number is scheduling. Stated target: from an average of 28 days to minutes once fully deployed. That figure measures scheduling cycle time. The wait for care itself is a different number. Still, it is an unusually concrete and auditable commitment.

My read:

Agentic AI in healthcare is moving out of demos and into the operating layer of America's largest integrated health system. The agents that scale first will be the ones that make the system run: intake, routing, verification, scheduling.