Presented by LMI & Carahsoft

Government agencies have spent decades collecting records they rarely had the resources to analyze. Maps, architectural drawings, contracts, and engineering documents contain valuable information, but extracting it has often required more time and money than an agency could justify. Artificial intelligence is changing that calculation, according to Alex Adamczyk, Vice President, Sales Engineering & Rapid Prototyping at LMI.

Speaking with Fed Gov Today at the Google Public Sector Summit, Adamczyk described a growing opportunity to apply AI to the government’s accumulated information. Much of that material is unstructured, and some remains on paper. Even when records have been digitized, finding useful details within them can still require painstaking review.

Recent advances in image analysis have expanded the kinds of documents AI can interpret, he said. Complex drawings and maps that presented difficult problems a year earlier are becoming more usable. For agencies, the opportunity begins with examining information they already possess and identifying a mission problem that information could help solve.

Adamczyk offered an example involving critical minerals. A customer concerned about mineral availability also held engineering drawings describing parts and materials in equipment routinely sent for disposal or recycling. Reviewing those drawings manually at sufficient scale had been economically impractical. As a result, potentially useful materials could leave the agency without being identified.

AI made it possible to examine the drawings for components containing rare earth magnets or other materials in constrained supply, he explained. That information could help the customer identify parts worth recovering before equipment entered the scrap stream. Records retained for years could support a different operational purpose without requiring a large team to inspect every drawing.

The example illustrates how AI can change the economics of an existing task. The underlying recovery opportunity was already understood. The obstacle was the expense of discovering which equipment contained the materials. Making that discovery less costly opened a path to acting on information the agency had long held.

Federal deployment introduces another set of challenges, particularly when sensitive or classified information is involved. Adamczyk said commercial technologies cannot simply be transferred into government environments without accounting for authorization requirements, cybersecurity controls, and the agency’s risk management processes.

He pointed to Model Context Protocol, which helps connect AI agents with tools and information, as an example of technology moving faster than familiar review practices. In the work he described, LMI’s cybersecurity team examined established approaches, including API gateway controls, and adapted them to address the behavior of AI agents.

That approach required explaining the proposed controls to government customers and preparing to adjust them as guidance developed. Adamczyk emphasized the value of federal experience in translating new capabilities into systems agencies can use responsibly. The mission benefit depends on making security part of implementation.

For agencies considering where to begin, the interview suggests looking closely at the records already on hand. A well-defined operational question can reveal information whose value was previously hidden by the cost of analysis. Connecting that question to appropriate technical controls is what turns an AI demonstration into a useful government capability.