Making AI Work in Manufacturing: 3 Industrial Leaders Share What’s Possible

Artificial intelligence has dominated conversations across manufacturing for the past several years. But for industrial automation professionals, the question is no longer, “Should we use AI?” Instead, it's becoming “Where does AI create measurable value, and where doesn't it?”

That practical mindset took center stage during a recent discussion at Automate Live, where experts from across robotics, machine vision, safety, and industrial software explored how manufacturers are moving beyond experimentation to real-world deployment.

Jimmy Carroll, host of the Manufacturing Matters Podcast, led the panel discussion live from Automate at McCormick Place in Chicago. Joining the conversation was Darcy Bachert, CEO of Prolucid Technologies; Charlie Long, VP and GM of Machine Vision and Fixed Industrial Scanning at Zebra Technologies; and Michele Silva, engineering manager at Reynolds & Moore.

The panel's message was clear: AI is becoming an indispensable tool for industrial automation, but success depends less on the technology itself and more on the data, infrastructure, and engineering decisions behind it.

Those themes will take center stage at A3's Advanced Vision & AI Conference this fall in Santa Clara, California, where manufacturing leaders, technology developers, and automation professionals will dive deeper into the technologies transforming industrial vision systems and AI-enabled manufacturing.

AI Is Becoming Practical

One of the strongest takeaways from the discussion was that AI is finally reaching a level of maturity where it can solve problems that traditional automation couldn't.

For years, industrial machine vision relied heavily on deterministic, rules-based programming. Engineers had to explicitly define every acceptable condition and every possible variation. While that approach remains highly effective for many applications, it struggles when variability increases. 

Modern AI changes that equation. Rather than programming every possible scenario, AI models can learn from examples and recognize patterns that would be difficult or impossible to define with conventional algorithms. That opens new opportunities for applications such as:

• Optical character recognition (OCR) with inconsistent fonts or damaged labels
• Complex visual inspections with natural variation
• Multi-modal inspection combining vision, temperature, vibration, and production data
• Autonomous robotic perception
• Intelligent quality control

Importantly, panelists emphasized that AI should augment, not replace, traditional vision tools. Simple measurement tasks and deterministic inspections are often still best handled using conventional machine vision algorithms, while AI fills the gaps where variability is highest.

Better AI Starts With Better Data

One recurring message throughout the discussion was that organizations often underestimate the importance of data preparation. AI isn't a magic solution that automatically fixes poor processes.

Successful deployments begin with structured, high-quality datasets and a clear understanding of the problem being solved. Without reliable data, even sophisticated AI models struggle to produce consistent results.

Manufacturers considering AI initiatives should first evaluate:

• Is the available data accurate and representative?
• Does the application truly require AI?
• >Can the model be trained with enough real-world examples?
• Where will inference occur — in the cloud or at the edge?

These foundational questions often determine whether an AI project succeeds long before model training begins.

Edge AI Is Enabling Real-Time Decisions

Industrial AI differs significantly from consumer AI. Manufacturing environments frequently require decisions in milliseconds, not seconds. Whether inspecting hundreds of parts per minute or enabling autonomous mobile robots (AMRs) to safely navigate dynamic environments, latency matters.

September 23-24, 2026 | Santa Clara, CA

Register

The panel highlighted why edge computing is becoming increasingly important for industrial AI. While cloud infrastructure remains valuable for training large models, inference often needs to happen directly on edge devices equipped with GPUs or specialized processors. Running AI locally enables systems to respond in real time without introducing communication delays that could affect productivity or safety.

AI Doesn't Replace Engineering Judgment

Another recurring theme was refreshingly practical. Companies shouldn't adopt AI simply because it's the latest technology. Instead, manufacturers should begin with the business problem.

If a straightforward rules-based solution already solves the application reliably, adding AI may introduce unnecessary complexity. AI delivers the greatest value when conventional approaches reach their limits. Panelists also cautioned against expecting AI to replace engineering expertise.

Documentation generation, quality management workflows, software development assistance, and safety management are all areas where AI can dramatically improve productivity. But experienced engineers remain essential for validating outputs, ensuring cybersecurity, maintaining regulatory compliance, and applying domain knowledge.

In manufacturing, AI performs best as an engineering assistant, not an engineering replacement.

Safety Remains Front and Center

As AI enables robots to work more closely with people, safety becomes even more critical. Panelists discussed how AI-powered perception systems are helping robots recognize people, navigate changing environments, and make safer operational decisions outside traditional fenced work cells.

At the same time, they stressed that standards governing AI-enabled safety systems continue to evolve, making careful validation and human oversight essential before deploying these technologies in production environments.

This balance between innovation and responsible deployment is becoming increasingly important as manufacturers explore collaborative robotics, autonomous mobile robots, and AI-driven automation.

Preparing for the Next Wave of Industrial AI

Perhaps the panel's most compelling message focused on people, not technology. The experts agreed that AI will likely reshape manufacturing more rapidly than previous technology shifts.

Rather than fearing that change, manufacturers should focus on developing new skills and learning how to work alongside AI tools. Engineers, technicians, programmers, and operations teams who understand how to evaluate, deploy, and manage AI will be well positioned as adoption accelerates across the industry.

Continue the Conversation at the Advanced Vision & AI Conference

These are exactly the types of conversations the Advanced Vision & AI Conference is designed to explore.

Bringing together experts from manufacturing, machine vision, robotics, AI, imaging, and industrial automation, the conference goes beyond discussing emerging technology to examine practical implementation, technical challenges, and lessons learned from real deployments. Attendees will learn from experts at Waymo, Intrinsic, Cognex, 3M, and more. (Stay tuned for more agenda details as they become available.) Register now. 

Whether you're evaluating your first AI vision project or scaling AI across multiple production lines, the conference offers an opportunity to learn from industry experts and connect with peers solving similar challenges.

As the panel made clear, the future of industrial AI won't be defined by hype. it will be shaped by thoughtful engineering, quality data, and practical implementation.