What does manufacturing’s ‘ChatGPT Moment’ mean for industry?

By Shinichiro “Shin” Nakamura 

According to experts, manufacturing is now entering its own “ChatGPT moment,” with this year marking an inflection point. Robot installations are expected to increase, and automation is gearing up for new roles on the factory floor. 

Manufacturers are no longer relying on artificial intelligence (AI) and automation only for report processing and insight generation. They’re now also increasingly using technology for sophisticated operations, with these tools set to become more autonomous and possibly even self-sufficient.

Plastics manufacturing, with its data richness and production intensity as one of the sector’s most in-demand industries, is set to lead the charge in early, large-scale transformation. Here’s what to keep in mind.

The forces behind booming adoption 

AI use in manufacturing is moving from theoretical to practical. Rather than simply monitoring data and insights, AI and automation tools are taking action in the form of robots and sensors that help control the factory environment. Deployment of robots in industrial settings is steadily growing.

Cost efficiency is one of the biggest driving forces behind mounting adoption, as well as operational gains. For instance, one manufacturer reports 12.5 percent material cost savings thanks to real-time adjustments from a machine learning-powered control system. 

Technology is also helping the life cycle of material come full circle in the value chain. To illustrate, sorting plastic for recycling has been a notoriously difficult area of operations. With smart sorting systems, significant labor and time burdens are eased.

Additionally, AI benefits manufacturers looking to easily procure readily available recycled materials that are somewhat time-consuming to source. That in turn strengthens environmental, social and governance (ESG) strategies by reducing carbon footprint, with reused materials more widely used in the procurement stage of the supply chain.

AI is stepping onto the factory floor 

Maintenance, productivity, health and safety, operational efficiency and real-time optimized material flow are just some of the key areas where AI’s impact is being felt in manufacturing. Strategies for AI deployment are rooted in three key workflow areas:

• Planning: Data-driven forecasting, scheduling and set-up optimization.

• Execution: AI-assisted operation and quality assurance on the floor. 

• Reporting and improvement: Feedback loops that inform continuous improvement and future planning. 

Generative AI (GenAI) is also advancing digital transformation from predictive maintenance and supply chain optimization to the rise of smart manufacturing ecosystems. This is powered by interconnectivity with other innovations such as IoT, machine learning and cloud computing. 

A budding partnership between leaders, operators, managers, engineers and AI systems is setting the tone for informed and proactive manufacturing production.

What true, scalable transformation demands 

Manufacturing environments involve complex practices, and needs can quickly change. Because of that, traditional models, when logic or calculation is clear, tend to be used for daily operations. GenAI, on the other hand, is used for exploratory purposes, applied by teams to refine or develop models or workflows. If the data feeding these tools is inaccurate, those inaccuracies or errors ricochet across operations. Moreover, GenAI on its own is not a predictive model. GenAI cannot be trained using small datasets or datasets that are lacking adequate data quality.

Most issues boil down to specific areas, and data is at the foundation of whether an AI deployment strategy will succeed or not. Worryingly, data infrastructure continues to be a serious problem for many organizations: almost two-thirds of organizations fail to keep track of their data.

With all the data to manage, it is not surprising that missing data, specifically around people, is one of the greatest challenges. Why? How people decide and react is extremely complex and obscure. Currently, manufacturers are trying to overcome this by seeing it in three parts: workflow, system and content. 

First, manufacturers must confirm existing workflows. From there, missing data, and how it affects wider strategic considerations, such as interoperability, safety and security, can be identified. New AI workflows should be designed by incorporating system and context to ensure consistent quality and safety adherence. This is a vital part of cementing human-machine collaboration.

Next is configuring system architecture. This includes gauging data-readiness across existing tech stacks, and whether systems are actually scalable within this digital environment. It also means ensuring that all systems are data-ready. Tools and systems that are not interoperable will stall, hallucinate and likely generate biased outputs or error-riddled insights.

Getting people on board 

Of course, the human factor cannot be ignored. Many automation strategies and AI integration attempts fail because humans are not kept in the loop. Change management is highly complex, particularly in the manufacturing world, where teams are large and scattered. Enforcing adoption and change management across the organization can be difficult because of that.

Manufacturers are also dealing with a lot of inertia, making it even more of a challenge to build momentum for change. The mentality of “why fix it if it’s not broken?” permeates the industry, which is known for sticking to manually intensive oversight for quality control and paper-based inputs.

This loops back to the workflow design step for scaling AI. Manufacturers need to validate existing processes before redesigning them to allow for a constant human-in-the-loop approach while facilitating real-time, automated insights. Focusing on workflow redesign with new system architecture and content helps bridge the gap between adoption and the capability of teams working alongside tools.

Hands-on, progressive training is another avenue to pursue. Arguably, this is the best way to help the workforce overcome fears and misconceptions of AI and its role in the industry while moving away from cultural inertia. Moreover, training is needed in strategies that should have user involvement in processes that check existing workflows, prepare new content and create new workflows. This reinforces the human-in-the-loop capabilities.

Weaving together these elements in AI deployment will keep manufacturers one step ahead. Enhanced integration built on interoperability that is paired with better data quality and informed, motivated teams is what will drive AI’s growing potential in predictive capabilities and maintenance.