Federal agencies continue to adopt AI technologies across research, mission-critical and operational spaces at an increasing rate. This trend is emphasized by the constant push for AI policies by federal leaders across the spectrum, proposed acquisition guidelines, and elements within the recently released FY 2027 budget proposal.

Moreover, GAO released a report earlier this week outlining key challenges in federal AI acquisitions. The report offers examples and recommendations aimed at improving the efficiency and effectiveness of AI procurement. These insights may influence federal decision-makers as they consider approaches for pursuing several of the AI-related initiatives proposed in the FY 2027 budget request.

Let’s dive into the GAO report first.

Challenges in Federal AI Acquisitions

At the request of Congress, GAO-26-107859 reviews the challenges facing federal AI acquisition, particularly given an increase in investment of the technology within the last 2 to 3 years. Findings of the report stem from interviews with acquisition officials at four agencies: DOD, DHS, GSA and VA. Challenges in federal AI acquisition reported by interviewed officials include:

• Lack of access to subject matter expertise: A limitation in the number of data scientists and software engineers carried barriers to defining requirements, understanding risks, establishing performance metrics and evaluating vendor proposals in AI acquisitions.
• Data and intellectual property (IP) rights protection: The balance between prioritizing and determining the appropriate level of data ownership with acquiring algorithms remains a constant challenge for federal agencies.
• Lengthy acquisition timelines and acquisition models: Traditional acquisition time frames and approaches do not align with the speed of AI development and evolution.
• Defining requirements and contract terms: AI acquisitions require precise and updated definitions of requirements and contract terms. Often, agency contract language standards are outdated or are not well defined to effectively evaluate AI performance.  
• Methods to test and evaluate AI technologies. Federal agencies face challenges in determining the appropriate approaches to test and evaluate AI before and after award phases due to the complexity of the systems.
• AI Pricing and overall cost: The challenge remains at agencies to completely understand AI costs. For example, federal agencies may underestimate overall cost for AI capabilities by failing to take into consideration costs such as cloud infrastructure or computing power.

The report notes that the April 2025 OMB Memorandum M-25-22 addresses several of these challenges. It also highlights DOD’s Maven program and GSA’s USAi acquisition as examples that demonstrate lessons learned and successful approaches.

For example, the Maven acquisition team developed a web of cross-functional teams of computer scientists, software engineers, cyber experts and more to support the procurement. Maven also used agile software acquisition approaches and 90-day sprints to accelerate delivery of capabilities and studied AI acquisition costs over time to help inform a comprehensive cost-estimate for the initiative.

At GSA, the USAi team developed privacy policy and contract language with clear data ownership expectations and limitations and used contract terms that were effective in previous AI acquisitions (i.e. contractual frameworks defining service expectations). The GSA team also reported testing multiple AI capabilities through a set of reliability tests before awarding multi-year USAi contracts.

Ultimately, GAO recommended that DOD, DHS, GSA, and VA update their AI policies to require the systematic collection and sharing of lessons learned from AI acquisitions. Incorporating examples such as DOD’s Maven and GSA’s USAi acquisitions could help agencies apply these insights to future AI initiatives.

Key AI Program Mentions in FY 2027 Agency Budget Requests

AI investments in the FY 2027 budget request are being maintained or expanded across several areas, reflecting the White House’s continued push to accelerate AI adoption. In October, OMB designated AI as a key FY 2027 R&D budget priority, emphasizing, “AI architectural advancements, high-performance AI techniques and systems, AI adversarial robustness and security, and the interpretability, controllability and steerability of AI systems.”

Investments from the budget see AI present in mission critical initiatives. For example, the FBI at DOJ is requesting $166M to allow the agency to develop AI-powered capabilities to respond to the impacts of global terrorism events by enhancing enterprise platforms supporting the Intelligence Community and leading law enforcement preparations for tactical response.

Moreover, agencies are investing in AI and machine learning in FY 2027 to improve administrative efficiency, particularly in areas of HR, financial management and regulatory review processes. Examples include:

• EPA: A $202M request to advance AI capabilities in areas such as rulemaking and permitting processes, categorizing public comments, developing communication products, and other programmatic needs.
• DOJ/EOIR: Seeks to leverage AI to modernize and automate efforts in four areas: AI Transcription of Immigration Proceedings, Modernizing EOIR's Case Management System, Improving EOIR's Applications, and Digitizing EOIR Records.
• OPM: Requests $2.5M within Workforce Policy and Innovation for AI and machine learning to modernize federal job classification and qualification standards across 22 occupational families and advance a merit-based, competency-focused Federal hiring process.
• Labor: Allocates $8.0M to scale and enable AI across the department, including meeting risk standards for high-impact use cases, scaling infrastructure for enterprise AI solutions, licensing, and initiating AI literacy programs.

The FY 2027 budget also reflects federal commitment to AI-driven analysis in national public health, research and disease surveillance. Examples include:

• HHS/FDA: Utilizing $2.0M to upgrade two centralized process systems with AI and ML capabilities to ultimately accelerate paperwork, enhance drug trials, and improve diagnostic accuracy.
• HHS/NIH: Describes harnessing AI to support the Real-World Data Platform and integrate data from electronic health records, wearable sensors, genomic data, and environmental exposures.
• HHS/CDC: Requests $45M in funding for the Biothreat Radar Initiative, which includes advances in pathogen genomics, laboratory diagnostics, data integration and AI-driven analytics to detect and track novel pathogens and early signs of emerging biological threats.
• VA/VHA: The agency is looking to utilize AI by leveraging VA data assets, including the Million Veteran Program, to support research in AI capabilities to improve diagnostic accuracy, optimize health care value, and develop personalized treatment protocols.
• VA: An additional $4.7M is requested towards VA’s AI infrastructure solution to enable the department to pilot and scale AI tools to improve operational efficiency, enhance clinical decision-making, and support personalized care and benefits delivery, while sustaining governance frameworks for safe, effective, and ethical use.

For further insight into the FY 2027 budget proposal, please refer to Deltek’s new report, FY 2027 Federal Budget Request: Priorities and Opportunities.