Army using 'agile' acquisition strategy as it moves forward on Project Linchpin

By Dan Schere  / September 11, 2023

As the Army begins to develop a pipeline for artificial intelligence and machine learning, service officials have described what they view as an "agile" acquisition process that aims to take advantage of an already-interested private sector.

Project Linchpin is the Army’s first initiative to create an AI and ML pipeline, and will initially focus on sensor systems such as satellites and optics, according to the service. The pipeline will initially support the Program Executive Office for Intelligence, Electronic Warfare and Sensors, but the “architecture and designs will be scalable beyond the PEO portfolio,” according to the Program Manager for Intelligence Systems and Analytics Office.

The Army describes Project Linchpin as being similar to software pipelines that are used on operating systems for cellphones, in which user data helps generate the next set of updates. The concept of an AI/ML pipeline for the Army would involve deploying those capabilities to intelligence, cyber and electronic warfare systems.

Since the Army started releasing requests for information related to Project Linchpin last November, there have been more than 170 industry engagements, the program manager’s office wrote in a Sept. 6 statement to Inside Defense. The service in particular wants to include small businesses and non-traditional defense contractors specializing in AI -- an area commercial industry is spending billions of dollars in.

“Additionally, we have learned this space evolves so rapidly that there is a need to develop a very agile acquisition strategy enabled by a rapid contracting approach that allows for maximum participation by industry. With the diverse and rapidly changing industrial base, and rapid evolution of technologies in this space, it emphasizes the need for our approach to be flexible and adaptable,” the program manager’s office said.

Bharat Patel, the project lead for Project Linchpin within PEO IEW&S, said Aug. 29 during a National Defense Industrial Association conference in Washington that he is working on an “OTA-type approach” over the next 18 months, referring to other transaction authorities.

“Right now, I'm not even worried about building models. I need to test out new infrastructure. I need to test my process, standards, all that. So, you're going to start seeing some OTA contracting activities,” he said at the conference.

When asked for elaboration on Patel’s comments, the program manager’s office wrote that the Army received feedback from 70 “industry partners involved in end-to-end ML ecosystem environments and companies with expertise and tools related to data holding/preparation, data labeling, model training, verification and validation, and deployment” in response to two RFIs released last December and in April.

The Army plans to begin awarding contracts for Project Linchpin in April 2024, and transition it to a program of record by 2026, Defense One reported in June.

Army Principal Deputy Secretary for Acquisition, Logistics and Technology Young Bang has emphasized that the service’s digital transformation must be guided by traceability, observability, replaceability and automated consumption, or TORC. According to the program manager’s office, industry is interested in "containerized methods for enforcing" TORC as well as improved transparency and “assurance of code, data, labels, labeling processes and telemetry capabilities” that will improve observation and tracking in ML ecosystems. In June, the Army released a broad agency announcement on the approach to TORC and expects to award contracts this month.

“We are continuing to refine the scope and strategy for contracts beyond the planned awards this month and will have more to share in the near future,” the program manager’s office wrote.

The Army Acquisition Support Center describes the approach to utilizing AI and ML on the battlefield through a scenario in which a soldier receives a sensor feed and uses AI to identify enemy equipment, as well as receive alerts about enemy movements. The soldier would be able to send messages to the commander, although there would still be a human in the loop.

Gen. James Rainey, the head of Army Futures Command, said during the AFCEA TechNet conference in Augusta, GA last month that when it comes to utilizing AI on the battlefield, the ability of the Army to realize its full potential means moving to a “genuinely data-centric approach to the network.” Adaptability, he said, is the characteristic that will be most key.

“Build a system that the number one characteristic is the ability to change and integrate technology as an opportunity as fast as it becomes available,” he said.

When asked by Inside Defense specifically whether the war in Ukraine has been informing the goals of Project Linchpin, the program manager’s office responded that the service is “leveraging all real-world and DOD exercises to inform the Project Linchpin strategy,” but did not refer to a specific conflict.

“We are collaborating closely with our operational users and requirements owners to identify relevant use cases and identifying workflows that would benefit from AI/ML,” the office wrote.