AI and the future of metal manufacturing

gorodenkoff/iStock/Getty Images Plus

I’ll admit it, ChatGPT scares me, but it also fills me with intrigue, even exuberance. All the fretting about artificial intelligence (AI) basically boils down to one question: Will it make us dumber or smarter? Probably both, but when it comes to business, only the smarter will thrive. The rest will offer products and services that will become commodities, utterly price-driven and transactional. Like machine automation, AI could be used to enhance our existing knowledge … or prevent us from needing to learn anything in the first place. The latter is the nightmare scenario.

High schoolers might have AI write a research paper, then run it through plagiarism and AI-checker websites to avoid getting caught. They do all they can to avoid actually learning how to write. Similarly, a shop owner employs press brake operators who learn to run the parts given to them, but that’s about it. They never learn about bend deductions and bend allowances, k-factors, how different metals behave, how the radius forms, backgauging strategies, bump bending—basically, all the possibilities air forming on a press brake offers. They just do what the machine tells them to do, and that’s it.

In these environments, companies might rely on technology as a crutch, implementing it so they don’t have to deal with people not showing up, people continually making mistakes, people not caring. Technology becomes the sole differentiator—until, that is, a shop down the street invests in similar equipment.

Now consider high schoolers who use AI not to avoid learning, but to enhance it. They know grammar. They know what makes an engaging sentence. They know how to build an argument and tell a story. They just use AI to build on what they already know. (Admittedly, that doesn’t sound like any high schooler I know, which is why AI is a bit scary.)

What makes AI so disruptive is how it could affect the knowledge jobs everyone thought were safe from being automated away. The reality might be a little more nuanced, especially when you consider how it could shape the future fab shop.

Consider again the operators in the press brake department. This time, the fabricator invests in training. They learn the grammar of metal fab and, with that foundation, know how to get creative and push the limits. They also start to connect the dots. They talk to laser and punch press personnel about microtabbing and nesting strategies, how several small brackets could be bent at once, how a mini-nest of small parts could be lifted and stacked by part removal automation.

The possibilities seem endless. In these environments, shop floor jobs are also knowledge jobs, and compared to other jobs in the fab shop, they’re less likely to be automated away. Even in the most heavily automated factories, people need to be there to maintain, run, and refine the process.

Move to the office, and the story changes. A few custom fabricators already offer automated quoting and order processing. Seven years ago, when I visited The Netherlands-based 247TailorSteel—one of the most impressive operations I’ve seen—I saw just two people in the office. Meanwhile, dozens of workers managed dozens of lasers and automatic-tool-change press brakes on the floor, along with a handful of automated guided vehicles that carried parts between processes. Customers uploaded 3D CAD files and received quotes instantly. Accepted orders were processed, scheduled, and nested automatically. A few shops stateside, including Utah-based OSH Cut (the company's founder Caleb Chamberlain now writes a column for thefabricator.com), are now following this model.

I recall the founder of 247TailorSteel saying he hesitated to offer additional services like welding and assembly, since automating the order flow would become more complicated. Quoting a cut and bent part is one thing; quoting a large subassembly, with a variety of fabricated and purchased components, is quite another.

Still, considering just how far computing power has come in recent years, will this always be the case? Probably not. And sure, the algorithms powering automated quoting, scheduling, and order processing aren’t in the realm of AI now, but they could be one day, especially considering the groundwork being laid in other areas. After all, AI now is writing software code. Google “AI code generator,” and you’ll get a list of websites that translate natural language into code. It might not be pristine, but it’s code nonetheless.

One day, we all may be able to just talk to an AI engine by saying, “Write a program that would track orders for this project.” Low-code apps like Microsoft Power Apps (which already comes with Office 365) are already being used by some shops to develop custom software, be it for job tracking, inspection reports, or anything else. Will low-code apps, with the help of AI, become no-code apps?

All this said, companies still need people to dream up such software in the first place. AI is still, well, artificial. It only does what we tell it to do, at least for now. If people use technology as a crutch, they’ll find it harder to dream up new possibilities. In an automated, AI-driven world, those possibilities, dreamt up by people with deep knowledge of metal manufacturing, will be what sets a company apart.