Two reports from The Defense Post, surfaced through Google News, point to the same shift: the US military is putting artificial intelligence to work on the unglamorous machinery of war rather than on the shooting end of it.
According to The Defense Post, the US military is testing an "embedded" AI training model for operational units — an approach that places AI instruction alongside the units that would actually use it, rather than treating it as a separate classroom exercise.
In a second report, The Defense Post says Gallatin AI has won a deal with the Defense Innovation Unit — the Pentagon office that buys commercial technology — to deploy what it describes as an agentic military logistics planner. "Agentic" is the industry's term for software that doesn't just answer questions but takes steps toward a goal on its own: in this case, working through the problem of moving people, fuel, ammunition and spare parts to the right place at the right time.
Both items are brief, and neither report as summarized here includes contract values, timelines or the units involved. What they share is a direction of travel. Logistics planning has historically been done by staff officers with spreadsheets and hard-won experience, and it is exactly the kind of constrained, high-variable problem where automated planning tools promise the biggest gains. Embedded training, meanwhile, addresses the other half of the equation — a tool nobody knows how to operate delivers nothing.
It matters because logistics and training decide whether militaries can actually sustain operations, so automating them would reshape how the US fights long before any AI is pointed at a target.