Operationalizing OMB M-24-18: A Practitioner’s Guide to Federal AI Acquisition

The Shift from Software to AI Capability Acquisition

Federal procurement is undergoing a fundamental shift as the Office of Management and Budget (OMB) moves to implement OMB Memorandum M-24-18, which builds upon the foundational governance established in OMB Memorandum M-24-10. For capture and proposal managers, this means the days of selling 'AI-enabled' software as a standard COTS product are ending. Agencies are now required to categorize AI acquisitions based on whether they are 'rights-impacting' or 'safety-impacting,' a distinction that triggers rigorous documentation and testing requirements.

Practitioners must understand that these requirements are not merely future goals; they are being integrated into solicitations now, with key deadlines for high-risk AI applications already in effect as of late 2024 and early 2025. The GAO-24-107332 report highlights that agencies are already implementing management and personnel requirements to meet these mandates.

The 'Show, Don't Tell' Procurement Model

One of the most significant changes introduced by M-24-18 is the move toward performance-based acquisition techniques. Instead of relying on written technical volumes that promise future performance, agencies are increasingly utilizing:

• Technical Trials and Prototypes: Contractors are required to demonstrate technical capabilities through live trials or proofs of concept during the evaluation phase.
• Oral Presentations: A shift away from static slide decks toward interactive sessions where technical leads must explain model provenance and security.
• Expert Consultants: Agencies are adding AI subject matter experts to Source Selection Evaluation Boards (SSEBs) to scrutinize vendor claims.

To win in this environment, capture teams must ensure their technical solutions are ready for immediate demonstration. This requires a 'lab-to-proposal' pipeline where the AI model's performance metrics are validated against the specific agency mission before the RFP is even released.

Navigating Rights-Impacting and Safety-Impacting Designations

Agencies must now disclose in solicitations if the planned AI use is rights-impacting (e.g., affecting civil liberties, privacy, or equal opportunity) or safety-impacting (e.g., affecting critical infrastructure or physical security). If your pursuit falls into these categories, your proposal must address:

1. Data Provenance and Quality: You must provide detailed documentation on the training data used, its source, and its appropriateness for the specific federal use case.
2. Ongoing Testing and Monitoring: Contracts will now delineate responsibilities for post-award monitoring to ensure the AI does not develop bias or performance degradation over time.
3. Risk Mitigation Protocols: Proposals must include specific criteria for risk mitigation, as outlined in the NIST AI Risk Management Framework.

Strategic Recommendations for BD Executives

First, audit your current AI offerings against the transparency requirements of M-24-18. Can you provide the 'trained model artifacts' for agency review if requested? Second, prioritize contracting vehicles that favor technical capability over Lowest Price Technically Acceptable (LPTA) models. The OMB guidance explicitly encourages agencies to avoid LPTA for complex AI acquisitions to ensure safety and reliability. Finally, monitor USAspending.gov for shifts in agency AI spending, specifically looking for increased funding in 'Other Transaction Authority' (OTA) agreements, which agencies are using to bypass traditional FAR-based delays for rapid AI deployment.