Recommended Free Tools
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
U.S. federal agencies buy AI tools by defining a mission need, researching available options, evaluating proposals, and setting contract terms for data, performance, and oversight. The process does not have one standard shape: an agency may seek a software product or an ongoing service, and an acquisition may use a contract or another agreement. Oversight continues after award, through monitoring and, eventually, contract closeout.
What counts as an AI acquisition?
An AI acquisition is not necessarily a purchase of a standalone product. The agency might acquire software it operates, or a service that supplies AI capabilities and outputs over time. GAO has also found that acquisitions can use contracts or other agreements, and may begin with an agency identifying a need or a vendor introducing a capability.
| Decision | What it changes |
|---|---|
| Agency-directed or vendor-introduced | An agency-directed acquisition starts with a mission need the agency has identified. A vendor-introduced capability starts with an option brought to the agency. Either route still calls for assessing whether the capability fits agency needs. |
| Contract or other agreement | Federal agencies use different agreement mechanisms. The route affects how the acquisition is structured; GAO’s report does not establish one mechanism as best for every AI purchase. |
| Software product or ongoing service | A product and a service can differ in how the agency accesses and operates the capability. The arrangement should make clear who is responsible for operation, support, data handling, and continued evaluation. |
These are choices to assess against the mission, the agency’s ability to evaluate and oversee the system, and the terms available for data rights, portability, and exit. This article concerns U.S. federal agencies; state, local, tribal, and territorial procurement rules are not addressed here.
How a federal AI acquisition proceeds
OMB Memorandum M-25-22, issued in April 2025, describes acquisition as a lifecycle: planning and market research, solicitation and evaluation, award, contract administration, oversight, and closeout. It calls for agencies to bring mission and technical perspectives into the process rather than treating AI procurement as a routine software purchase.
1. Define the mission and likely uses
The agency identifies the problem the acquisition is meant to address, the intended users, and foreseeable use cases. OMB calls for cross-functional teams and consideration of whether the planned use could be high impact. That early work gives the agency a basis for deciding what capabilities it actually needs and what risks it will need to assess.
2. Research the market and test claims
Market research helps the agency compare available capabilities and approaches, including new entrants where appropriate. OMB recommends demonstrations and testing in realistic operating conditions when practicable. Testing proposed solutions against the agency’s needs can reveal limitations that a vendor presentation or specification alone may not show. Planning should also account for long-term costs and the difficulty of changing suppliers.
3. Set evaluation criteria and solicit proposals
The solicitation should explain relevant aspects of the use context and what documentation the agency needs. Agencies should establish measurable outcomes and assess proposals against them. Where practicable, testing should help the agency understand both what a proposed system can do and where it may fall short; a vendor’s general performance claim is not a substitute for evidence relevant to the agency’s intended use.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems4. Award with terms the agency can enforce
Contract terms should translate the agency’s requirements into obligations it can monitor. OMB highlights data use, privacy, intellectual-property and licensing rights, pricing transparency, portability, knowledge transfer, and protections against vendor lock-in. It also recommends preserving appropriate rights to code or models produced under contract and access to components needed to operate and monitor a system.
5. Operate, monitor, and close out
Before deploying an AI system on an agency-operated information system, the agency needs the required authorization to operate. After deployment, contract terms should support recurring assessment of performance, risks, and effectiveness, including independent evaluation using agency-defined data where applicable. Agencies should assess the system’s value and continuing costs, and plan for closeout, including transfer of data and derived assets when the contract ends.
What data and contract terms deserve particular attention?
Data rights are central to whether an agency can control information it supplies, understand what the vendor retains, and move to another solution. OMB says agencies should establish processes for data ownership and intellectual-property rights, including the scope of licenses. Contracts should limit vendor collection and retention to what is reasonably necessary to perform the contract.
Rank #3
For non-public agency inputs and outputs, M-25-22 says contracts should prohibit use to further train publicly or commercially available AI unless the agency explicitly consents, consistent with applicable law. This is a contract safeguard, not a claim that every AI vendor or system automatically follows the same practice.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →To make a future change of provider more feasible, agencies can address knowledge transfer, clear data and model portability, transparent licensing and pricing, and appropriate rights to code or models developed under contract. These provisions matter most when considered before award: an agency that lacks access to necessary data, components, or operating knowledge may have fewer practical options later.
How agencies can oversee AI after award
Oversight is not limited to checking whether a tool met its initial specifications. OMB calls for ongoing assessment of performance, risks, and effectiveness, and for monitoring privacy, civil-rights, and civil-liberties risks. Agencies should establish, where practicable, criteria for ending use if a system no longer meets its needs or presents unacceptable risks.
GAO’s accountability framework offers four complementary questions for procurement and operations:
- Governance: Is the system’s purpose clear, and are responsibility and decision-making authority assigned?
- Data: Is the data appropriate for the use, and is it handled with suitable quality and safeguards?
- Performance: Is there evidence that the system meets the agency’s intended needs?
- Monitoring: Can the agency detect changes in performance or harms over time?
Because system inputs and operations may not always be visible to an agency, third-party assessment and audit can be relevant parts of oversight.
What GAO has found about federal AI buying
In a report published April 13, 2026, GAO said federal agencies’ AI use more than doubled from 2023 to 2024. That finding describes reported use, not the number of acquisitions. GAO examined 13 acquisitions at the Department of Defense, Department of Homeland Security, General Services Administration, and Department of Veterans Affairs, and found varied acquisition approaches through fiscal year 2025.
The report also identified practical obstacles: agencies described difficulty accessing technical experts to evaluate proposals and understanding AI-related costs. GAO found that the selected agencies were not systematically collecting and sharing acquisition lessons. It recommended that DOD, DHS, GSA, and VA establish policies requiring systematic collection and submission of lessons to a GSA-managed repository. The agencies concurred, but GAO listed the recommendations as open in the report.
That finding points to a broader implementation challenge: a well-designed acquisition depends on people and processes able to evaluate technical claims, estimate costs, and use experience from previous purchases. It is distinct from the government-wide policy baseline. In December 2023, GAO reported that OMB had not yet issued government-wide guidance on acquiring and using AI; OMB issued M-25-22 in April 2025.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
