VA Approves Sidekick AI to Help Veterans Navigate Benefits and Support

The U.S. Department of Veterans Affairs has approved Sidekick, a conversational AI platform designed to help veterans navigate healthcare, housing, employment and benefits. USC disclosed the approval on October 9, 2026, following months of privacy, safety and accuracy review. The platform pairs information selected by VA specialists with a route to human support when individual judgment is needed.

Sidekick was co-developed by USC Research Professor Skip Rizzo, director of medical virtual reality at the university’s Institute for Creative Technologies. USC describes the veteran-focused service as the platform’s first implementation, bringing his work on virtual humans into a conversational interface for finding resources and understanding next steps.

Specialists set the information; AI handles the conversation

The starting point is a defined collection of material: thousands of pages selected by 10 VA subject-matter experts. Those specialists cover areas including healthcare, employment, housing and the GI Bill. The service is intended to help veterans navigate education and benefits alongside those other needs.

USC describes a deliberate division of responsibility. Subject-matter experts determine the information the platform uses, the subjects it covers and the boundaries of its role. AI then organizes and retrieves that knowledge, handles the interaction and makes the material available through conversation.

Veterans can ask questions in ordinary language and continue the exchange toward relevant resources, applications and courses of action. The intended task goes beyond locating a document: USC frames the problem as helping someone understand how a resource applies to their circumstances, which form they need and what to do next.

A navigation role, not an autonomous expert

The platform’s stated boundary is important for questions involving personal circumstances. Sidekick is not intended to become an autonomous expert. When an issue requires individual judgment, it can direct the user to appropriate human support, rather than keeping every part of the exchange within the AI system.

Rizzo presents that arrangement as a way to fill gaps in support, not replace workers. He says there are not enough people in those roles to provide all the assistance veterans need.

The human boundary predates Sidekick

Sidekick builds on nearly three decades of Rizzo’s virtual-human research at USC. An earlier project, SimCoach, gave service members and veterans a private interface in 2009 for discussing mental-health concerns and finding routes to care. It encouraged clinical help when needed rather than attempting diagnosis or therapy.

In 2013, SimSensei added real-time analysis of facial, vocal and other behavioral signals associated with psychological distress. More recently, Battle Buddy brought patient-facing support to mobile devices, helping veterans develop coping skills, make plans for staying safe during crises and remain connected to their support networks.

Across those projects, Rizzo and his collaborators had to define what virtual humans could responsibly do and where responsibility belonged with a person. The team developed and documented approaches to privacy, safety, clinical scope and human oversight. Rizzo argues that this earlier work supplies a framework for using modern AI at greater scale with ethical safeguards.