We Taught the System to Think. Now What?
From anomaly detection to autonomous code generation, Elbit Systems engineers are facing a new reality – AI isn’t just accelerating their work, it’s beginning to shape it.
What happens when software begins to write itself and machines start making manufacturing decisions independently? At Elbit Systems, these aren’t theoretical questions anymore. From production floors to development centers and data labs, AI is no longer just a support tool – it’s an active partner in designing, testing, and optimizing advanced defense systems.
In a recent Innovation Week webinar, Elbit engineers and executives, Uri and Avichay as well as well-known Israeli podcaster Ran Levi discussed how generative tools, machine learning, and real-time analytics are transforming how the company develops both software and hardware. The focus wasn’t on replacing humans – but on redefining where human expertise matters most, and where intelligent systems can help to eliminate bottlenecks, boost precision, and detect hidden flaws before they turn into real issues.
When the Code Becomes a Black Box
“We’re entering an era where people are building software without understanding what’s inside,” warned Ran, referring to the rise of ‘vibe coding’ – AI-generated software whose internal logic is hidden from users. “You get working systems, but you don’t know how they work.” That might be fine in gaming or entertainment. But in defense? That’s a red line.
Avichay, from Elbit’s C4I&Cyber division, agreed. “I expect Microsoft Copilot and generative tools to help me write code faster and cleaner,” he said. “In those mission-critical systems, we will use such tools to support decision making, keeping a man-in-the-loop strategy for now.”
For now, Elbit is testing AI-enhanced development tools like automated unit testing and simulation engines.
When AI Joins the Factory Floor
Some of AI’s biggest gains at Elbit have been on the production floor. At the company’s small arms facility, an AI-powered system recently enhanced manual inspections of 5.56 ammunition rounds. The outcome? Faster throughput, more consistent results, and the removal of a major production bottleneck.
“We’re seeing efficiency improve by dozens of percentage points,” said Uri. “Machines are detecting faults before they trigger shutdowns – analyzing vibrations and spotting visual anomalies that humans wouldn’t catch in time.”
This isn’t automation for automation’s sake – it’s about capacity. “Demand is pushing our physical limits,” Uri said. “We’re using AI not just to maintain quality, but to keep up.”
Uri also shared that Elbit is working with Israeli startups to automate CNC code generation – one of the more obscure but important parts of mechanical manufacturing. “That alone is clearing a major bottleneck.”
From Code to Combat
What happens when AI leaves the factory floor and joins the fight? Avichay offered a look at the battlefield ahead. “Modern intelligence gathering puts an enormous strain on analysts – hours of screen time, stitching together events from endless data feeds. AI can ease that load. Generative systems can help find the needle in the haystack.”
He described one example using LLM-based agents – AI systems that work together like a team – to analyze complex surveillance footage, flag likely anomalies, and provide decision-support insights. “It might not give me exactly what I need,” he said, “but it gets me much closer, much faster.”
Ran raised a deeper question: How does AI function amid battlefield ambiguity – when information is incomplete, misleading, or constantly shifting? Avichay pointed to the idea of human-machine teaming. “AI agents can move through buildings, urban areas, dynamic environments – but they still can’t make judgment calls on their own. The machine will know what to recommend but it is the human who will make the final decision. It’s the human-AI partnership that makes the system resilient.”
More Than a Tech Race
Elbit Systems isn’t pursuing AI for the buzz. Its approach is rooted in operational reality: identify real-world challenges, adapt the tools accordingly, and always strive to keeping the end user in focus. That includes new ways to share data with customers, AI-driven quality control across high-demand supply chains, and internal innovation calls – offering a funded R&D program for employee-led AI initiatives.
In a world where machines increasingly think, Elbit is making a more deliberate decision: systems that support human decisions rather than override them. It’s not just about building smarter systems – it’s about building smarter operations.