The federal acquisition environment is undergoing several changes, among them, is the use of AI in procurement-related activities. The federal contracting community is utilizing AI in a variety of settings, including pre-award administrative tasks such as drafting requirements, checking solicitation compliance, validating forms, and conducting market research.

The inclusion of AI in federal procurement is emphasized in an August 2025 GSA posting, which sought industry input on utilizing AI to supplement and improve the federal acquisition lifecycle.  

Agencies are also increasingly piloting and using AI within the proposal source selection process. For example, GSA previously reported implementation of the Contract Acquisition Lifecycle Intelligence (CALI) tool. ACT-IAC reports that CALI is an automated machine learning evaluation tool which assesses vendor proposals against solicitation requirements to determine compliance and aid in source selection.

Additionally, the Army issued an SBIR topic last year called, “AI Enabled Source Selection Solution for Contract Proposal Evaluation” seeking a tool to, “address principal challenges experienced during the evaluation and source selection processes including length of time required to conduct evaluations; varying levels of experience possessed by evaluators; inconsistent treatment of Offerors by evaluators; and deviation from solicitation instructions and evaluation criteria, potentially leading to protest.”

The benefits of AI in source selection span helping evaluators manage workload, identify key insights, and bring consistency to complex reviews. However, the downfalls in using AI in source selection include errors in output (i.e. hallucinations) and the risk of absence in human judgement in critical phases of proposal evaluation. Regardless, contractors are beginning to see government language in solicitation postings alluding to the employment of AI as a tool to assist proposal analysis and review.

In a lively discussion on the use of AI in source selection within yesterday’s event hosted by the Baroni Center, “AI in Proposal Evaluations: What Could Go Wrong?,” panelists addressed how to use AI to accelerate procurement processes while balancing trust and averting risks. Primary participants of panel included Matt Colantino, Product Manager of Quanitfy at AlphaSix; Kevin Misener, Senior Attorney at HHS; and David Timm, Partner at Burr & Forman, LLP.

Nearly all panelists agreed that humans must remain in the loop and government acquisition officials should not solely rely on large language models (LLMs) to land a decision. Rather, LLMs are good for bringing together analysis and highlighting key insights and should provide all the documented evidence required to make the decision, said Colantino.

Misener stated that benchmarking to establish the AI tools used in source selection and the ability to have independent mechanisms to run them internally will instill additional confidence in the technology.

Moreover, Misener suggested that building the AI tool with the agency data itself, though people in government must be aware of the limitations of pre-award data (for example, historical data may not reflect recent changes, and some websites are not optimized enough for AI to pull from). Thus, government must put controls in place to make certain data accessible.

Colantino added that solicitation formation is also very critical to the success of AI in source selection, adding that this is where benchmarking is important.

So, where do contractors go from here?

The growing use of AI in source selection seems inevitable, and contractors should take several steps to prepare for this evolution. Timm offered key pieces of advice to contractors, including asking in RFI and pre-bid settings questions to the government on if AI will be used in the source selection process. Timm also stated that asking HOW AI will be used and WHICH model will be used would also be worthwhile for contractors. Knowing which model will be used will help in proposal preparation as AI solutions tend to use their own model data the most, Timm explained.

The Burr & Forman partner also suggested that contractors request an oral briefing with government to gage how much human involvement was involved in the evaluation. Timm also suggested that contractors consider protesting if it appears that there was no human in the loop within the decision. Protests set the standard for best practices and hold government accountable, stated Timm.

In proposal preparation, contractors are also advised to use solicitation terminology and section headings, include information in text and not only graphics, and audit proposals before submission, among others, according to an April 2026 Federal News Network article.

The question of AI in source selection is no longer if, but when and how. As AI technologies settle into the federal procurement process, both agencies and contractors face a period of adjustment. Agencies must invest in ensuring AI tools produce reliable and defensible results. Contractors, meanwhile, must prepare proposals with AI analysis in mind. In this environment, preparation is no longer just about writing a strong proposal, but rather understanding the AI environment in which that proposal will first be read.