RallyPropRallyProp.ai
ARM26BX06-NV012ActiveSBIR

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

Department of DefenseARMY

AI Overview

This RFP seeks a schema-driven AI system that transforms complex acquisition decisions into auditable, traceable decision packages for ground vehicle programs. The prototype will demonstrate how agentic AI processes digital engineering data while maintaining human control, transparency, and bias awareness throughout multi-phase trade studies.

This summary is AI-generated from the official solicitation.

Key Details

Agency
Department of Defense
Funding Amount
—
Release Date
September 2, 2026
Due Date
October 21, 2026

Official Description

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 En...

View on official source

Change History

Q&A UpdatedOct 15, 2026 at 6:01 PM

Material change · Requirements + Scope + Description

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

  • Q&A set substantially rewritten. The prior version had nine questions, several marked with "Response Pending" style gaps. The updated version has fourteen questions, with most earlier answers replaced by firmer ones.
  • New topics added: proof-by-demonstration scope (Q2), whether the system may flag insufficient evidence (Q2), provenance from recommendations back to sources (Q12), human override with preserved machine output (Q11), and SysML v2 interface standards (Q8).
  • Commercial AI services now explicitly allowed for Phase I: commercial cloud-hosted foundation-model APIs may be used on unclassified synthetic or public non-CUI data (Q3, Q9). CMMC Level 1 applies; no CUI or classified data in commercial APIs.
  • Phase I deliverable clarified: contractor-hosted demonstration with documentation is acceptable; packaging for the GVSC engineering network is not required until Phase II (Q10).
  • Signer role still unnamed: no PM, PEO, MDA, or PAE is required as signer; design for a generic acquisition decision authority (Q5, Q13).
  • Bias scope confirmed: decision-framing bias is the primary Phase I object; statistical and model-output bias are also in scope (Q1, Q6).
  • Success criteria left to offerors: no ranking of time savings versus auditability versus analyst effort has been published (Q7).
  • Live-case and GFI status: Phase I scenarios and models must be offeror-developed; PAE guidance is for Phase II only (Q4).
Q&A UpdatedOct 15, 2026 at 5:01 PM

Material change · Requirements + Scope + Attachment + Description

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

The Q&A content changed, but the visible text shows no substantive change to the answers. Comparing the previous and updated versions, Q1 through Q9 and their answers appear identical, and the Q9 text is cut off in both versions. The only apparent difference is in the truncated tail of Q9 ("Does DEVCOM-GVSC or PAE curren..."), which cannot be compared because both versions end at the same point.

Applicant impact: none identifiable from the visible text. If the change is in the truncated portion of Q9 or in content not shown, the full Q&A should be checked for new questions, answers replacing "Response Pending" entries, or removed items, since those would matter for eligibility, deliverables, or submission rules.

Q&A UpdatedOct 13, 2026 at 7:03 PM

Material change · Requirements + Scope + Description + Submission

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

  • Q&A set expanded and rewritten: The previous version had 13 mostly open, "PAE guidance expected" style questions. The updated version restructures the Q&A around new questions, with several previously pending answers now filled in.
  • Phase I API/ICD (new Q1): Phase II is now confirmed as the first required interface milestone; a Phase I API/ICD artifact is useful but not required. Self-contained Phase I is acceptable.
  • Conflicts of interest (new Q2): Participation does not automatically bar a vendor from later bidding on related electronics or vehicle-network work, but OCI is determined by the Contracting Officer under FAR 9.5/DFARS. Specific facts should be raised with the CO.
  • Deployment and bias scope (new Q3): No air-gap requirement or on-premises compute limits have been published, and hosted model services are not prohibited. Phase I bias checks cover decision-framing/cognitive bias primarily, with statistical/model-output bias in scope. No signer role or Decision Package template is mandated.
  • PAE data timing (Q4, Q5): PAE guidance is planned for Phase II only. Phase I can finish on offeror-synthetic scenarios. The final Phase I demonstration format (live, recorded, or report) remains unconfirmed.
  • Single-copy deliverable (Q6): The form is not specified and is primarily an end-of-Phase II requirement. Phase I software need not run on the GVSC engineering network.
  • LLM use (Q7): Commercial cloud and commercial LLM APIs are acceptable for synthetic or public, unclassified, non-CUI data; no Government-authorized cloud is mandated for Phase I.
  • Refresh triggers and output (Q8): No refresh trigger is ranked; a readable document plus a machine-readable record is suitable.
Q&A UpdatedOct 13, 2026 at 6:02 PM

Material change · Requirements + Description

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

  • New Q5 added (schema-level interoperability): Asks whether Phase I requires demonstrated interoperability with a specific digital engineering export format (SysML v2 XMI) or a digital-thread/twin repository. Answer: schema-level interoperability against representative artifacts is sufficient for Phase I feasibility; live integration with SysML v2 XMI or an existing repository is not required; a self-contained demo environment is acceptable; SysML/UML alignment is useful, and DE MIL-HDBK-539A is referenced as optional best practice.
  • Renumbering: Former Q5 through Q14 shifted to Q6 through Q13+, so question numbers changed but the substance of the DAOSoft, data-source, and open-source questions is unchanged.
  • Former Q9 (CUI/cloud safeguarding) and Q10 (open source, non-CUI, non-classified) answers: Wording is unchanged apart from renumbering.

Why it matters: Applicants no longer need to plan for live SysML v2 or digital-thread integration in Phase I. A self-contained prototype with representative DE artifact ingestion, mapping, and traceability is enough, which lowers Phase I scope and cost.

Q&A UpdatedSep 25, 2026 at 3:01 PM

Informational change · Description

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

Description information changed.

Status ChangedSep 23, 2026 at 1:03 PM

Informational change · Status

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

Status changed from status: Pre-Release to status: Open.

Q&A UpdatedSep 18, 2026 at 1:02 PM

Informational change · Description

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

Description information changed.

Opportunity AddedSep 2, 2026 at 12:02 PM

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

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

Similar Opportunities

Get Alerts for Opportunities Like This

RallyProp monitors thousands of funding sources and sends instant alerts when opportunities match your technology and capabilities.

Start Free Trial