Agentic-AI, Schema-Driven Decision Management for Auditable Studies and Acquisition Decisions

This topic will prototype and demonstrate a schema-driven Decision Management (DM) foundation plus a governed agentic AI layer that turns messy decision requests into auditable, signer-ready Decision Packages. The focus is proof by demonstration (not a full product build) to show the approach works for point trade studies and long-horizon acquisition decision programs with traceability, robustness, and human control. AI can process vast amounts of structured and unstructured data from Digital Engineering (DE) ecosystems, including digital threads and digital twins, enabling comprehensive analysis across the entire product lifecycle. Interoperability across models is achieved as AI integrates data from various DE tools (e.g., SysML 2.0, UML, etc.) and provides a unified view of system models, enabling cross-functional teams to collaborate effectively. The first objective is to demonstrate a unified representation that supports both (a) single-episode trade studies and (b) multi-episode "decision programs" with refresh cycles. This includes implementing context engineering specifically for the ground vehicle domain through knowledge graphs and semantic layers that capture domain-specific relationships, constraints, and requirements. The second objective is to implement agentic capabilities that produce structured artifacts and reproducible outputs, while maintaining human control and predictable behavior per established human-AI interaction guidance. AI will automate the validation of technical documents, models, and outputs, ensuring compliance with standards and requirements throughout the decision-making process. The third objective is to extend the schema so bias checks, risks, and assumptions are explicit, testable objects – so decision readiness can be assessed before sign-off and refreshed responsibly when conditions change. The system will leverage semantic understanding to maintain consistency across digital engineering artifacts and enable traceable decision rationale.