The drone threat facing Israel is no longer simply Iranian. Iran gave Moscow the Shahed concept; Russia then subjected it to years of war, learning what survives electronic warfare, penetrates air defense systems and can be manufactured cheaply at scale. The latest jet-powered Geran variants now fly at more than 300 miles per hour, carry larger payloads and are proving considerably harder for Ukraine to intercept. What should worry Israel is not only the weapon, but the speed of the learning cycle behind it.
That cycle is moving back toward the Middle East. Russia has begun sharing with Iran modified Shahed technology and tactical lessons learned in Ukraine, turning what was once largely a one-way transfer into a two-way military relationship. Israel may increasingly face systems designed in Iran, improved under Russian battlefield conditions and returned to Tehran with better engines, navigation and resistance to jamming. The next Shahed attacking Israel may embody lessons learned over Ukraine.
This matters because the economics favor the attacker. Shaheds are not cutting-edge technologys, and that is precisely the point. They are cheap, simple enough to produce in volume and capable of forcing sophisticated air defenses to react. Their numbers can consume stocks of vastly more expensive interceptor missiles. Mix drones with ballistic missiles, cruise missiles and decoys, and even an excellent defense faces a magazine, cost and decision-time problem.
Israel has built one of the world's most sophisticated layered air-defense systems. But no defense has infinite interceptors, crews or radar attention. A cheap attacker does not need to be better than an Israeli or American interceptor costing hundreds of thousands or millions. It only needs to force that exchange repeatedly. Ukraine has learned that the fight is not simply missile against drone. It is an industrial competition over who can adapt, manufacture and replenish faster.
This is where Israel should think about artificial intelligence differently. Most military discussion of AI focuses on integrating artificial intelligence into the weapon: autonomous navigation, target recognition, swarming and electronic warfare. Israel should also use AI inside the machinery that decides what to buy. Finance is already considering AI agents that continuously search for better returns and move customers toward better alternatives. Defense procurement needs its own version - not a machine spending autonomously, but an agent matching battlefield requirements with technology, production capacity, price and delivery speed.
Imagine Israeli intelligence determines that Iran is preparing a faster generation of one-way attack drones. An AI procurement agent should immediately ask which current interceptors can catch them, which sensors need software updates, which electronic-warfare systems can defeat them more cheaply and which Israeli companies already have relevant prototypes. It should search allied experience, compare combat data and determine who can produce 5,000 units in three months rather than 500 a year. A commander would still decide, but the difference is that their decision could be informed in days rather than months, and based on hard data.
Ukraine has already demonstrated part of this philosophy. Its war economy increasingly links operational units with a fast-moving drone industry, while battlefield feedback influences what gets modified and ordered next. Europe, by contrast, is trying to rearm through procurement institutions built for another era. The problem is not simply spending: Europe needs to move faster toward drones, electronic warfare, military AI and agile suppliers capable of much faster innovation cycles rather than preserve acquisition processes around technologies that may change before contracts are signed.
Israel is unusually well suited to close that gap. Soldiers, engineers, startups and major defense companies operate in close proximity, and years of conflict have generated something more valuable than another demonstration: operational data. Israel knows which systems detected targets in time, which interceptors worked, which equipment survived jamming, which suppliers delivered and which prototypes failed. An AI procurement layer could connect those lessons with inventories, manufacturing capacity and emerging threats, creating institutional memory that updates continuously.
This would not eliminate human judgment or the politics of procurement. Sensitive data would still require robust cybersecurity, safeguards would be needed to prevent manufacturers from manipulating rankings, and officials would remain accountable for final procurement decisions. AI also cannot overcome fundamental capability gaps when the necessary technology simply does not exist. But it can reveal capability gaps earlier, identify viable alternatives faster, flag production bottlenecks, and prevent bureaucracies from spending months rediscovering information that already exists elsewhere in the system.
Israel’s challenge, therefore, is not simply to outbuild Iran, but to outlearn an Iran-Russia military ecosystem that is adapting rapidly. The concern cannot stop with Iran: cheaper, proven technologies tend to migrate to proxies and partners. The Shahed's evolution into faster Russian Geran variants is a warning about the shrinking time between threat and countermeasure.
In the drone era, superiority may belong less to the country with the most sophisticated platform than to the country that can observe, decide, procure, manufacture and adapt fastest. Israel already possesses exceptional engineers, operational experience and defense companies. What it needs is a procurement architecture capable of learning at something approaching the speed of the battlefield.
The next decisive use of AI in defense may therefore not be the algorithm flying the drone. It may be the agent watching the battlefield, understanding what has changed, searching industry for the answer and telling the military what to buy before the adversary changes again. Against cheap weapons modified every few months and launched by the hundreds, that difference in time can matter as much as the interceptor itself. The next AI weapon may not fly. It may be the buyer.