Edge AI Brings Faster Decision-Making to Federal Missions

Imagine a future where an autonomous drone files over a battlefield, identifying enemy combatants, scanning for wounded personnel and determining the best evacuation route to extract them.

Meanwhile, AI agents are assessing injuries and advising medics on potential care options, while monitoring what treatments have been applied. All of this could transpire in terrain that offers little connectivity or network support.

AI has dominated the discussion around modernizing federal agency operations for years, but the technology’s potential to make operations more efficient far from the data center could be its biggest coup.

While many agencies are looking to the potential of productivity gains in their back-office functions or customer-facing services, some are exploring capturing real-time gains at the field level by utilizing AI models to speed compute directly from devices at the edge of the network, better known as edge AI.

By shifting compute from out of the cloud and having it processed directly on the device, agencies can see faster results in missions including national security, space operations and homeland security, and as a result, can perform faster decision-making.

“It used to be that you were waiting six to seven hours to get an updated picture from a satellite image or something like that. Now, you are getting a detection — not just a picture, but a full detection with enumerated background and everything else — in minutes,” said War Department Chief Digital & Artificial Intelligence Officer Cameron Stanley at the Billington Cybersecurity conference on Sept. 8.

Because edge AI can be deployed on a range of devices — from sensors and unmanned systems to satellites and airport security screeners — the applications for agencies’ use are wide-ranging and the gains are only beginning.

How Edge AI Works

Rather than collecting data and sending it to a cloud network to be processed, edge AI does its data processing at the source by equipping the device collecting the data with AI models that can run the compute necessary on site.

The advantage of this process is that agencies are able to both free up network resources for jobs requiring larger compute and receive information faster, reducing latency. Edge AI capabilities also operate in low network-connection environments, allowing users to continue to operate while sending the data back to the network once connection is restored.

But for them to work, it requires both sizable datasets for the AI to train on and a partnership with industry to develop the AI algorithms needed to power the edge devices.

Agencies are tackling the challenges in a number of different approaches, both in how they are deploying edge AI capabilities and how they are contracting for them.

What It Takes to Bring AI to the Battlefield

The Telemedicine and Advanced Technology Research Center (TATRC) — the technology innovation arm of the U.S. Army Medical Research and Development Command (USAMRDC) — looked at how automation could improve battlefield care in future conflicts. Specifically, the team wanted to preserve the 60-minute window, or “golden hour”, to move a person from an injury site to the next point of care and increase survivability.

“During the past 20 years, we were able to have amazing survivability on the battlefield because of something called the golden hour,” said Matt Quinn, TATRC’s science director, at the Sept. 22 ACT-IAC Digital Transformation Summit. “That was because we controlled the sky; it was relatively small amount of distance to cover. We just can’t assure that, and so one of our approaches to address this is being able to address the mass casualty and prolonged care through automation, helping those frontline caregivers do more and do it for a longer period of time.”

Because of the potential for edge AI capabilities, Quinn said TATRC looked at automating identifying injuries at a mass casualty event, tracking the care provided by medic and logging the data on medical resources used, which he said accounts for 25% of the time medics spend on patient care.

“If you just offload manual data collection, that’s 25% more capacity that we could have,” Quinn told GovCIO Media & Research.

The challenge was training AI algorithms to recognize potential casualties in austere conditions from which little to no training data had previously existed.

Quinn said TATRC created a training dataset of simulated casualty events and leveraged an other transaction authority (OTA) contract competition to receive AI algorithms from industry that the center could test for potential prototypes.

The Vision and Intelligence Systems for Medical Teaming Applications (VISTA) system deployed two technologies in 2025 for testing, its Point of Injury capability to identify casualties with drones and its Patient Handling Automated Recording and Observation System to monitor patient care.

“In actually a relatively short amount of time, we were able to get from nothing basically to some basic level of identification of the state of patients, the resources used and the actions of the medic. And with additional funding and additional time, we could do a lot more,” he said.

While these technologies are still in the pilot and testing stages, Quinn added that edge AI capabilities exist today and would be critical to any future solutions moving forward.

“Our context of use is that we are not going to be connected and if there is a solution that requires connection, it’s just not feasible in the kind of environment both for security reasons, but also just for time,” he said.

How Agencies are Looking to Use Edge AI

Because of its ability to more efficiently share certain types of data quickly, edge AI has already seen several use cases develop across the federal government. Agencies are testing and deploying edge AI tools across use cases that could range from forest fire prevention to satellite navigation.

The Pentagon most notably views edge AI capabilities as an expansive scope of applications to address various missions because of the technology’s data-processing ability in disconnected, disrupted, intermittent or limited bandwidth (D-DIL) environments where network coverage is constrained.

The use cases for edge AI in D-DIL environments range from supporting forward units equipped with their own edge devices to planes and unmanned aerial systems leveraging it to process inputs faster.

Space Force has explored edge AI as a way to help improve the data compute onboard its satellites as a way to increase space domain awareness (SDA) and improve processing times.

Edge AI touches multiple components of agency mission within the Department of Homeland Security, most notably in its research and development arm, DHS Science & Technology Directorate.

DHS S&T has developed programs to capitalize on the data from edge AI sensors to better detect and predict the path of wildfires, partnering with the University of California San Diego to develop a platform to help firefighters collect and analyze fire data in real-time and use models to project where it’s headed next.

NASA officials foresee AI applications deployed to a wealth of mission functions, helping speed up operations like data downlinking while already deploying it plan safe travel routes for the Perseverance Mars rover.

While use cases continue to be developed and tested across multiple agencies, edge AI’s capacity as a platform will continue to facilitate these new applications.