There is an odd paradox when you stroll through the maintenance hangar at a facility like Utah’s Hill Air Force Base. The aircraft have a futuristic appearance. It appears to be the future, with the F-35s parked in their bays, the panels open, and the technicians wearing flight suits holding tablets. However, up until recently, the paperwork looked a lot like the past. clipboards. notes written by hand. Decades’ worth of maintenance logs, written by retired airmen. The AI story quietly starts with that mess of handwritten history.
Since 2017, C3 AI has collaborated with the Air Force’s Rapid Sustainment Office on a platform known as PANDA, or Predictive Analytics and Decision Assistant. It sounds like a PowerPoint slide, and it most likely began that way. However, the function is actually helpful. From contemporary sensor telemetry on an F-35 to the natural language processing of those faded handwritten notes on older airframes, PANDA consumes both structured and unstructured data in an attempt to provide an answer to a single question that has plagued fleet operators throughout history: when is this thing going to break?
PANDA became the Air Force’s official system of record for predictive maintenance in 2023. Although it’s a bureaucratic title, it’s important. It indicates that the program is no longer in pilot status. It is a backbone. One of the officers in charge of the project, Lt. Col. Michael Lasher, has publicly discussed the enormous amount of data involved, including millions of maintenance and supply records, years’ worth of jet telemetry, and sensor streams that are practically impossible for a human team to analyze. It’s not that the AI is more intelligent than a mechanic. The idea is that a mechanic cannot even see all of the data.
Investors are beginning to take an interest in the financial narrative that lies beneath this. According to C3 AI, which had a difficult few years as a publicly traded company, the DOD could save up to $5 billion a year if predictive maintenance were fully implemented. It’s a significant number even if the actual amount is only half of that. Organizations using AI-based predictive maintenance report downtime reductions of 35 to 45 percent and maintenance cost savings of 10 to 25 percent, according to more general industry data. These percentages translate into jets flying rather than sitting for the Pentagon.
The way this is subtly changing contractor margins is intriguing. With a backlog of almost $190 billion, GE Aerospace has turned to digital twins and AI-driven engine diagnostics. The major players, including Raytheon, L3Harris, and Lockheed, have integrated predictive analytics into their platforms. It’s possible that the legacy defense companies who can incorporate these features into the systems they already sell will end up being the true long-term winners rather than the flashy AI names. When service revenue becomes recurring, margins increase, and predictive maintenance is fundamentally a uniform-wearing recurring-revenue engine.
As this develops, there’s a sense that the Pentagon has stumbled into something remarkably useful. The majority of defense AI discussions quickly veer into cyber, autonomous vehicles, and drones. This one is more modest. It’s about getting a jet through one more flight before something breaks. Years ago, the Defense Innovation Unit identified the cost savings. Similar programs for helicopters have been developed by the Joint AI Center. Army aviation, the Navy, and the Marines have all been called in.
It’s still unclear if the entire $5 billion will come to pass. Between fiscal years, government initiatives tend to lose steam. However, the course has been decided. Jets who were accustomed to waiting for failure are gradually learning to raise their hands before it occurs.