The Trump administration’s new memo on what agencies should consider when buying artificial intelligence capabilities neither breaks new ground nor rehashes previous guidance.
Rather, the Office of Management and Budget is recognizing the constant state of change in how agencies acquire and use AI.
Kraig Conrad, CEO of the National Contract Management Association, said the memo stands out for its emphasis on oversight during the post-award phase of federal contracts.
“In any contractual relationship anywhere, whether it’s public procurement, federal or commercial, the post-award is where dollars are destroyed,” said Conrad in an interview with Federal News Network. “Anything you thought you negotiated in the front end goes away in performance. So that conversation of having a regular check in, even with less people possibly doing this work, which means they’ve got a few more things that they now have to do. But giving some clarity of the expectation post award, I think is powerful. I think having a standing reporting requirement, or at least timing, every five months or every quarter, about how the project is doing, I think that’s a good thing to put some discipline around that. Hopefully that means there’s someone there able to receive this information and actually act, because again, it’s in the post-award phase where the destruction of the contract value occurs.”
OMB laid out only four deadlines total in the memo, two of which are for the General Services Administration and two are for all agencies.
GSA must develop a plan to release a guide publicly to assist agencies in buying AI tools within 100 days.
It then has 200 days to create an online portal only for federal agencies for knowledge sharing, such as sample solicitations or standard contract clauses and negotiated costs for common AI systems and other relevant documents.
Agencies have two main deadlines. One is in 270 days when they must update, where necessary, existing internal procedures on acquisition to comply with the memo. This includes reviewing planned acquisitions involving an AI system or service and provide any feedback on AI performance and risk management practices. It also means they should convene a cross-functional team to help with the coordination and decision-making processes associated with the acquisition of AI, and they must ensure use of contract terms for intellectual property (IP) rights.
In 200 days, agencies have to make sure they have processes and contractual terms to address the use of government data and clearly delineate the respective ownership and IP rights of the government and the contractor. OMB says agencies should give careful consideration of respective IP licensing rights, especially as an agency procures an AI system or service where agency information is used to train, fine tune and develop the AI system.
Implementation questions remain
Conrad and other federal acquisition experts gave the memo better than average marks overall, but said there still are a lot of questions that OMB needs to answer.
One industry executive, who requested anonymity because they didn’t get permission to speak to the press, said a lot of these deadlines are in fiscal 2026 and OMB doesn’t address any AI bought this fiscal year.
“Some of these initial orders are hard to figure out and a lot is left up to the individual agencies to figure out, so that will cause them to struggle with being consistent. This one references lot of different things, but by targeting three initial areas, it at least helps frame the universe,” the executive said. “With technology that moves this quickly, you will run into problems. I don’t think you can get away from that by just issuing a new memo.”
Experts said OMB’s new memo takes some of what the Biden administration outlined across 36 pages last September. For example, both the new and old memo tell agencies to avoid vendor lock-in when buying AI tools. Another example is a focus on data portability across AI systems and ensuring the government continues to own data and IP rights to systems.
John Ferry, the president Trenchant Analytics, which builds acquisition AI tools for agencies, said he was pleased to see the continued focus on how agencies can avoid vendor lock-in when buying AI tools.
“At a more granular level that speaks to experiences that government has had in the past with implementing data platforms or implementing applications that take government data, do some sort of modification to it, and based off of off the shelf IP rights that are written contracts, the outputs from that processing then becomes some level of unique company protected data,” Ferry said in an interview. “There are a number of companies in the market, frankly, that kind of have made a lot of money that way.”
He added that agencies struggle with getting the data right for these AI tools so that’s going to be a big challenge before they can really focus on the IP and data rights issue.
Focus on contracting approaches
One big difference in the memos, is the Trump administration’s memo goes further by calling out specific types of contracts to use when buying AI tools and capabilities. In this case, agencies should focus on statements of objectives and performance based contracts with incentives, instead of firm fixed price type of contracts.
NCMA’s Conrad said that approach may take some getting used to as agencies don’t necessarily buy technology this way.
“They’re talking about being more agile. They’re talking about something that establishes a buy that’s a learning buy, meaning it’s going to iterate. I think this is a good thing to hear about that, particularly if your statement of work implies that it’s going to learn. And if there is a better way to do it, then that comes back into the process,” he said. “So yes, it might be more expensive, but hopefully through this process of getting frequent updates from the community of suppliers, that community that help inform the acquisition teams to what’s new, what’s changing so they can make better buying decisions when those new features come up again. I think the drive here is do not contract for a solution that you think is 100% for the next 10 years. It’s more about how do we find places where we can be agile and learn about a new features.”
Ferry and other experts also said the move toward different contracting approaches may be a big lift for several agencies.
The industry executive said beyond traditional Defense Department weapons systems, agencies have used incentive fees and award fee type contracts.
“They are difficult to explain to auditors and you get dinged by audit agencies because everyone gets 100%, a lot like the contractor performance and assessment rating system (CPARS),” the executive said. “It’s a wildly complex activity that the government will struggle with, especially as contracting officers and program managers are being offered to leave government. This is especially true for those who are at the end of their careers and may have done this type of work in the past.”
Additionally, the executive said agencies will also struggle to find the right balance with using a SOO approach that could let them add new capabilities every quarter or 30 days.
“It will be a challenge to administer those types of contracts,” the executive said.
AI to relieve contracting burdens
A lot of the answers to these challenges will come from the implementation guidance agencies issue.
Ferry said the memo doesn’t address underlying technology and talent issues, nor does it give agencies enough specifics around mitigating risks and securing the data.
This is especially concerning with the memo’s focus on agencies adopting high impact AI systems —which Ferry said — may be a mistake as an initial focus and instead agencies should look at another type of use case.
“To introduce AI and introduce capabilities to government, really the best place to start is actually [things that are] easy to implement, take a low amount of time to implement, but will have an impact on a lot of people or have a high impact on a few number of people,” he said. “The reason for that is, with any new capability, call it organizational change management, adoption is the point of the realm. If someone’s not using a new capability it’s absolutely worthless, and every dollar you spend on it is also wasted.”
Another key piece of the memo is using AI tools in acquisition itself to make the processes less burdensome. OMB is telling agencies to apply AI for things like market research. The Homeland Security Department, for example, led a pilot using AI for market research and for contract close out efforts.
GSA also has started testing generative AI tools for acquisition to help contracting officers and other acquisition workers search across all acquisition policies and internal directives/circulars, as well as code generation.
Ferry said AI for acquisition processes are a perfect match because of the linear nature of the processes and the fact it’s fairly standardized across government.
“In terms of a macro use case for AI in government, acquisition is something that affects everybody. It is an ideal use case for the acquisition workforce, which is under a lot of pressure because a lot of folks have left. The budgets are only going up. The number of 1102s is only going down. So that delta is a major gap so there is a strong need,” he said. “The workforce is also proactive. They want to adopt AI, and this is not like in the DoD where there are far fewer ethical questions about using AI for dropping bombs than there is using AI for cutting contracts. So you have that confluence of a user community, a use case and an application that are ideal.”
Copyright
© 2025 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.