Kostiantyn Gitko is the CEO of Devox Software, specializing in AI-native transformation and scaling complex systems for global enterprises.
getty
In Part 1, I showed how combining industrial IoT and AI helps manufacturers achieve the level of resilience required today. In Part 2, I will examine what drives true resilience in modern manufacturing.
Today, companies are adopting IIoT to monitor operations and using AI to adjust production execution in real time based on machine signals. These use cases are a major focus at industry events, where AI-powered chips guide robotics and integrated cameras track movement. Applied AI uses statistical models to fine-tune high-precision machinery, enabling tight control through continuous feedback analysis. But the global focus tends to be on new technologies, while critical knowledge about legacy equipment is at risk of being lost.
Today, manufacturers are facing the "silver tsunami"—a wave of retirements among a generation of master technicians whose decades of experience shaped deep operational intuition. Manufacturing leaders are using AI and computer vision to basically "download" a veteran’s intuition before they walk out the door.
But at this stage of maturity, it’s never just about the tech. It’s about a new organizational model where AI-in-the-loop safeguards the most valuable asset: expertise. This genuinely futuristic process starts with digitizing paper blueprints and handwritten notes.
Next, pros use smart glasses or work within a camera setup as AI analyzes their techniques, including how they handle tools, the angle used to tighten components and their step-by-step workflow. The system records all spoken commentary, and NLP models convert unstructured speech and technical jargon into clear instructions. All data runs through an RAG architecture with knowledge graphs to ensure accuracy and consistency.
This makes the internal AI model a reliable source of truth and a training platform for new technicians. Gen Z workers, who often dislike lengthy rule books, can just ask questions aloud via a tablet or smart device, and the system uses advanced ML to recognize their voice—even in a noisy factory environment.
In an instant, the system generates concise instructions or plays a video clip of a veteran technician explaining the process. For example, Siemens has developed AI co-pilots for industrial operations that can help workers generate, interpret and translate technical information and troubleshooting guidance in natural language. Augmentir tailors content to each operator’s skill level to speed onboarding, and BAE Systems has developed augmented reality display systems that overlay visual guidance and operational information into a user’s field of view for industrial and defense applications.
Companies can start deploying this technology on production lines today, thanks to custom AI model development. The core architecture is relatively simple: IIoT cameras and an AI framework capture every action and commentary from the master technician. The AI then translates these inputs into clear language, preserving the knowledge as structured video, audio and text. Beyond recorded one-off workshops, these IIoT systems can also run alongside legacy equipment, capturing experts’ daily interactions on the fly.
This raises a related question: Where can we find specialists equipped to work with this new generation of manufacturing equipment and cobots? I see this opportunity in the Gen Z cohort, who are digital natives at their core.
Just as critical knowledge shouldn’t be locked in one or two heads, even within a unified digital system, intelligence must be distributed closer to where decisions are made. It's balancing computing workloads across the manufacturing floor.
The "cloud-first" mantra sounds great in a boardroom, but in reality, it can be a recipe for latency. On one hand, digital systems continuously collect and process data, with modern industrial equipment generating telemetry at very high rates—up to 100 kilohertz. Sending this volume of data to the cloud requires significant resources.
On the other hand, AI as an overarching layer requires dedicated compute, robust data pipelines and MLOps to sustain continuous processing. At the same time, every system in modern manufacturing operates within specific parameters regarding energy consumption and network bandwidth.
Increasingly, companies are shifting their focus from moving data to strategically distributing intelligence across a hybrid architecture. They optimize these environments by sizing compute appropriately for the models, ensuring network consistency and managing energy distribution effectively.
How does it work? You can think of this edge system as a digital nervous system that reacts to "pain" before the central system registers the anomaly. During mass production, the system actively scans for micro-deviations from this benchmark. As soon as tool wear causes even minor metric changes, the edge system acts like an experienced technician. It instantly sends a deterministic command to the machine controller, slowing the feed or adjusting parameters and correcting the process well before a defect can form.
To build this kind of system, follow these core principles.
Install industrial gateways directly next to the equipment for localized processing of high-frequency data. The algorithm captures the "baseline state," mapping every vibration and thermal shift. Once the system starts flagging micro-deviations before they become defects, it is ready to act like an experienced technician.
Configure the system to scan for the slightest variances from the benchmark during mass production. At the first sign of tool wear, the edge system sends a deterministic command to the controller, correcting parameters before defects form.
Design the digital ecosystem with anticipated operational tiers and straightforward recovery algorithms. Integrate manual processes as a reliable safety net and a rapid, localized method for resolving rare incidents.
Over time, even the smartest systems require optimization to streamline components and consolidate logic.
Applying these principles keeps systems adaptable across environments. Streamlining the infrastructure improves stability during operational shifts. For this reason, systems are designed with multiple operational modes for different environments, favoring adaptable frameworks.
In the third and final installment, I plan to highlight the key metrics for a successful AI and IIoT rollout in manufacturing. In my view, and in the view of our manufacturing partners, resilience is now a key competitive factor and extends beyond engineering, driving core business strategy.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。