NATO is handing the cheap work in war to AI. The irreversible work stays human.

Militaries are starting to sort which parts of war to hand to machines by a simple test: what is cheap and abundant, and what is irreversible. NATO's Eastern Flank Deterrence Initiative is the cleanest public statement of that test so far.

The doctrine is neither "AI pulls the trigger" nor "AI stays out of war." It is asymmetric allocation. The cheap, plentiful work is flying thousands of drones along the Russia-Belarus border from northern Finland to the Black Sea, plus the sensor and satellite plumbing that turns them into one network. The scarce, irreversible work is deciding which human being dies. NATO hands the first job to machines and anchors the second with people. US Army Maj. Ben Schiff, the primary architect of the EFDI Data Backbone, has put the reframe in plain words: the alliance wants to "reverse the concept" of autonomy in war. Most readers will read it as reassurance or a slide toward killer robots. That reading misreads the intent, Schiff suggests. It treats flight, loitering, and sensor fusion as a workload problem, the kill decision as a moral one. Ukraine has been running this experiment by force for two years, and the alliance is now naming it as policy.

The mechanism travels: when a workload becomes cheap, the scarce resource shifts, and doctrine follows. The remaining questions are whether a human "on the loop" can keep pace with a swarm, and whether every ally accepts the same rule. The seam between declared policy and operational reality is the part to watch.