AI drives metal fabrication forward

Chor muang/iStock/Getty Images Plus

The business of custom metal fabrication boils down to having the right information at the right time. This applies to every step of every business opportunity—not only throughout each order cycle (initial quote to the final shipment), but how those orders change over time. Good information makes fabricators more competitive, and in the coming years, artificial intelligence (AI) will prove just how powerful good information can be.

Software overall has changed the way custom fabricators collect, process, and analyze data. Consider advances in enterprise resource planning (ERP) platforms, especially in their ability to communicate and integrate with machines and external software programs, like nesting and shipping software, to lower costs and increase efficiency. AI takes these capabilities to new levels.

The competitive advantages that arise will push more importance on gathering the right data, be it in customer relationship management (CRM) systems, job quoting, job tracking, or time spent converting customer drawings into usable formats. All this will give AI the foundation custom fabricators need to truly change how jobs flow through the business and, ultimately, how they succeed in the market. Keeping manufacturing technology current keeps fabricators in the game, but those who make best use of information will win.

How Sales Will Change

Imagine a salesperson’s typical workday. They check on regular accounts; research key market areas; reach out to existing and new prospects; and meet with managers, quoting, and estimating personnel about new and potential work.

Effective salespeople manage a complex web of external and internal information. One missed call, email, or customer note made in CRM software can be incredibly costly. Competitive intelligence, perhaps recorded in a quarterly report, SEC filing, or general news source, can be incredibly beneficial. The quickly evolving large language models (LLM) of AI will be able to capture it all and boil the information down into actionable steps.

LLMs can review correspondence with customers and prospects, identify changes, and aggregate CRM data. Supporting this will be internal data from everywhere in the business: inventory, delivery, and quality of raw materials (flatness for laser cutting, consistent thickness and chemical properties for forming, etc.); win rates for certain kinds of jobs in certain sectors; estimated versus actual job costs as reported by ERP and job tracking; and more. All this helps pinpoint a fabricator’s true strengths and weaknesses, which gives sales and top managers alike the information they need to sustain and grow the business.

As LLMs evolve and become more intelligent, they will help salespeople record and collect information, such as insights from a recent phone call or a report from a news feed. LLMs will detect and correct data-entry issues, like extra zeros or a misplaced decimal point (not just for sales but for everyone in the shop).

They will also analyze current scenarios, like a loss of a customer, compare them to past events, and make educated suggestions on what to do differently. Was the customer loss due to price, reliability, or both? And what does that loss say about the customer, the sector that customer plays in, and how it all fits within a fabricator’s overall business strategy?

Quoting, Pricing Strategies, and Forecasting

Anyone who knows the classic “wastes” in lean manufacturing—defects, excessive processing, waiting, transportation, and more—learns how to identify which activities add value and which do not. But what about quoting? Traditionally, shops might use bid-win rates to determine what to quote. If a fabricator rarely wins work from specific customers, a shop might think twice about spending a lot of time quoting their work. That said, by not quoting certain work, fabricators might be missing some significant opportunities.

Here, AI, together with the Industrial Internet of Things (IIoT) and advances in ERP platforms, could be truly transformative. Today, quoting platforms already automate much of the process for certain jobs, and such software-enabled automation will likely become even more sophisticated in the years to come. No longer will CAD technicians need to extract and interpret components from a SolidWorks file manually. Software is getting better at extracting the right information automatically, even from PDFs.

Data from IIoT platforms can give quoting personnel real-world machine setup, run-time, and material handling data. This can complement the clock-in and clock-out data from ERP, as well as material data. Consistent material might cost more, but less setup and processing time might pay for increased material quality many times over—if, that is, the data supports this assumption.

In the age of AI, data is king, so making information easy to record will be more important than ever. Such information can narrow the gap between estimated and actual job costs. The narrower the gap, the more competitive and strategic a shop can be with its pricing. ERP has helped fabricators do this for years, and AI is set to take this to the next level.

For instance, AI’s deep data processing will allow fabricators to model how customers will respond to price changes based on historical sales data. AI predictions aren’t 100% accurate, but they inform gut feelings about effective pricing strategies.

Similarly, AI will be used to predict demand for products based on historical data, market trends, and customer behavior. All this could help optimize production schedules, shorten lead times, and avoid stockouts.

Production, Quality, Training, and Maintenance

A pallet of blanks arrives at a press brake, at which point the work ticket is scanned, showing the job is now ready and staged for bending. Once the operator is ready, he or she scans the work ticket again. Depending on the job and the shop, operators might scan work tickets or travelers to record setup times, job run-times, and material handling times. In shops that have implemented IIoT technology, machines can report such information directly, but operators still might need to clock in and out of jobs manually.

Machine run-times matter, but in most custom fab shops, people still need to move parts to and from those work centers, and the time for that work needs to be accounted for. Cut parts don’t shake themselves out of sheet skeletons; they don’t present themselves to forming, welding, or painting; and they don’t package themselves for shipment.

Again, in the age of AI, shops that collect and use the right data will win. AI can examine disparate jobs and sequence everything for optimal throughput. That’s a tall order for fabricators that run hundreds of different jobs at any one time—juggling schedules for purchased components and outside services like plating and heat treating.

But given the right datasets, AI will be up to the task. What’s the true cost of nesting different jobs on a sheet, looking a certain number of days ahead in the schedule? What’s the true cost of moving and maintaining work-in-process (WIP)? What’s the true cost of scheduling disparate jobs back to back (more setup time) versus grouping like jobs together (more WIP, if parts aren’t immediately needed downstream)? A seemingly small change can send ripples of variability up and down the value stream, and given the right information and enough historical data, AI could help planners choose the best way forward.

Of course, what about all those scans, clicks, and manual data entry tasks? Most shop managers want their operators producing parts, not tediously reporting tasks. Here, AI could evolve to help on several fronts. First, because AI helps turn vast amounts of data into actionable information, the act of reporting job times will yield greater benefits, as AI helps planners organize and release jobs in new, better ways. The more front-line workers record what they do, the easier their jobs become.

Second, AI will make the act of reporting and data-gathering much more automated. Today, RFID and related technologies can track where jobs are; when an operator retrieves work, the system can detect and record those tasks, initiating “job start” times automatically. AI-enabled vision technology might make such automated tracking even more flexible and seamless. One day, cameras might constantly scan a shop for WIP and jobs being processed. In this world, operators won’t need to scan anything as AI captures it all by video, feeding that information back to a comprehensive system. It will know which parts from a nest were sorted when, where specific sheet remnants are stored, and how long it took to move material from one point to another.

All this dovetails into quality assurance. Today, a shop might record the scrap it produces, but the reasons why the scrap was produced might not be fully captured. Why did setup at the press brake take so long? Why did certain pieces fail a weld inspection? AI could look at a mountain of nonconformance reports (NCRs), then synthesize the information to provide guidance for future improvements. How do current issues relate to past NCRs for similar jobs? An operation might not have fabricated anything similar since, say, a small job five years ago that no one remembers—but AI does not forget.

AI doesn’t forget machine shutdowns either, be they planned or unplanned. The technology will elevate predictive, sensor-based maintenance and use historical data to refine preventive maintenance schedules. For example, if a specific kind of cutting job puts excessive slag on the table slats, the system will know that the slats need to be cleaned sooner rather than later, and even account for that in production and capacity planning.

All of the above relates to labor management and training. When production planners account for capacity that’s truly available, and refined maintenance practices make unexpected shutdowns a rarity, everyone’s job becomes more predictable. Employees work in a proactive environment. Instead of putting out fires, they prevent them from sparking. Training is baked into the work schedule—with, again, AI guiding who needs to be trained when, and even how (accounting for one’s style of learning, the need for repetition, etc.).

Inventory and Purchasing

Working with the latest ERP platforms, companies have developed key metrics that help them make informed decisions when it comes to what inventory to keep and when. But those metrics won’t be able to compete with AI-driven tools that can tap into real-time data. What types of material should be kept in stock, and when, to account for customer demand as well as manufacturability requirements? What material can be delivered in hours, and what has significant lead time? AI will answer those questions and more, instantaneously.

AI won’t just draw from data within the shop, either. A major publicly traded customer might publish growth predictions for certain products for the upcoming quarter. How will this affect other links in the supply chain, or the business of outside service providers like custom coaters or heat treaters? Geopolitical events might affect supply in myriad ways. What about inclement weather? How do these trends relate to past events at similar times of the year? AI will take all this into account and more.

Everything All at Once

Managing a busy custom fab shop can be overwhelming. A client calls with last-minute changes. Another customer won’t return a call to clarify an order. A machine operator doesn’t show up, and the scrambling ensues. Managers can feel they need to look everywhere at once, and even then, surprises emerge.

The problem, of course, is that humans simply can’t look everywhere at once. A shop might fabricate an order halfway, then get an unexpected call: A supply chain problem threw a wrench into production. Acting in good faith for a longtime customer, the fabricator holds the inventory and waits. It’s just one of those things, right?

Not necessarily. AI could have looked at historical data and determined that supply chain problems with this customer or industrial sector aren’t unusual. That information could have changed how schedulers chose to release orders to the floor. Sure, the problem might be unavoidable, but armed with AI, planners might also find a workaround.

Humans simply can’t look at years’ worth of order histories with every job and compare that to machine and labor capacity, current and past job and customer mixes, demand trends, maintenance issues, supply chains, geopolitical issues, and on and on. They can’t look at everything at once—but AI can.

AI won’t make decisions, but it can make people better decision-makers. Like other technologies, AI will eliminate some jobs, but it won’t eliminate the need for good people. To the contrary, AI will make good people even more valuable than they already are.