Home / Past Performance / GRACE AI
Case study · ARPA-H

GRACE AI: secure generative AI inside a federal boundary.

A secure generative AI assistant that showed how cutting-edge AI can be deployed within strict government security frameworks to accelerate research and enhance productivity.

Gen AI
Deployed in a federal boundary
Zero
Security incidents
95%
User adoption
$2.5B
Research portfolio supported

The challenge

ARPA-H needed to give researchers the productivity of modern generative AI without stepping outside the security and compliance frameworks that govern federal research data.

Off-the-shelf AI tools could not meet the agency's requirements for data handling, access control, and auditability. The agency needed an assistant that lived inside its security boundary and met federal standards from the first line of code.

Our approach

DAWNE designed and delivered GRACE AI, a secure generative AI assistant built to operate entirely within the agency's approved security environment. Security and compliance were foundational, not bolted on. We paired the assistant with Microsoft Power Platform applications and modern ITSM so the agency could adopt, manage, and extend it with confidence.

  • Generative AI architecture aligned to federal security frameworks
  • Identity, access control, and auditability built in
  • Microsoft Power Platform applications for everyday workflows
  • ServiceNow and Zendesk ITSM for support and operations
  • Training and change management to drive adoption

The outcome

GRACE AI demonstrated that innovation and compliance are not a trade-off. The assistant accelerated research workflows and enhanced productivity while maintaining zero security incidents, supporting a research portfolio valued at $2.5B with strong user adoption.

The project exemplifies DAWNE's commitment to delivering innovation without compromising security, compliance, or operational integrity.

Want AI your agency can actually deploy?

We can help you move from pilot to production, securely.

Request a briefing →