This article is part of The Wisconsin Idea, an investigative reporting initiative focused on rural Wisconsin.
On the outskirts of Mellen, Wisconsin, just across the Bad River, sits North Country Lumber — a sawmill yard full of neat, lego-brick buildings.
Despite the high demand for lumber, Mellen is a small town growing smaller. Mirroring trends across the state, its younger workforce has been steadily relocating to industrial hubs like Milwaukee or Green Bay, which offer a wider variety of job opportunities.
Jon Koosmann started working at North Country Lumber two days after graduating high school in 1990. At first, he says, “it was just something for me to do to make some money.” After a stint at a nearby veneer plant, he returned to the lumberyard and now serves as the mill’s hardwood grader. It’s a job that in a few years’ time may no longer exist — or, at least, not in its current form.
Wood grading — the process of sorting wood by quality for different uses —is an essential job in any lumber mill. However, as stagnant wages and alternative job options push graders from the industry and more mills begin to automate, this task is increasingly falling to a new type of employee.
Artificially intelligent computers are faster than human labor — if not necessarily more accurate — and are quickly becoming popular with big corporate enterprises that can easily afford cutting-edge equipment. Tech manufacturers are capitalizing on this trend by touting sleek new AI products, creating an economic feedback loop.
Situated on Menominee and Ojibwe ancestral land — and home to over 10 indigenous tribes — Wisconsin became a U.S. state in 1848. As white colonists migrated to the area and railroads expanded, demand for home-building timber skyrocketed. Vast swaths of wilderness suddenly became an invaluable commodity, and, seemingly overnight, Wisconsin’s lumber economy was born.
Early lumbermen worked without much machinery. “Technology was oxen and horses and sleighs, an axe and a cross cut saw,” says Jerry Apps, a Wisconsin-based rural sociologist. “And yet, it was successful and got the job done.”
Trees were cut by hand, placed into major rivers and floated downstream. Sawmills had to be located along a river in order to harness the water’s power to saw the lumber into usable boards.
The job looks much different today. Apps, who owns a tree farm in Waushara County, has watched automation take root over the last few decades. “You would never even recognize it, really. Because the work the loggers do on my farm, my gosh, they sit inside of an air conditioned camp with a computer screen in front of them,” he says.
Wisconsin’s wood industry has stagnated in the last two decades; from 2003 to 2008, about 20% of sawmills in the state closed, with an increase of about 14% in the following five years.
Forty years ago, Mellen was a bustling town of over 1,000; today its population has dwindled to less than 650. One by one, local bakeries and bars shuttered for good. “It’s the same old story,” says Jeff Delegan, a National Hardwood Lumber Association (NHLA)-certified wood grader and president of North Country Lumber. “Little shops were eaten up by Amazon or Walmart … And I think it’s a lot more noticeable when you’re in a small setting like this.”
Delegan attributes falling interest in industry jobs to the town’s general decline.
“Fifteen years ago, I probably had ten to fifteen applications on hand at all times,” Delegan says. Nowadays, he says he’s lucky to get one every three months.
Other mills report similar figures. A 2018 Wisconsin DNR survey showed that 36% of wood manufacturers claimed they were often or always short on labor, predominantly in sawmills. An additional 32% reported that they were sometimes understaffed. The lack of cushion can lead to major headaches. Workers have to be shuffled around to compensate for a sick or absent employee, spreading the staff thin. “It’s like robbing Peter to pay Paul,” says Koosmann.
While Mellen’s ebbing population likely accounts for some of the labor shortage, workers point to stagnant wages and the physically demanding nature of the job to explain low entry into the profession.
“It’s slowly dying out. It’s like, you know, Cadillac car owners,” says Rado Gazo, a professor of Wood Processing and Industrial Engineering at Purdue University. “It’s a demanding job; you have to really like what you’re doing… It’s just, people have other options.”
Koosmann and his coworkers clock an average of forty-five hours a week. They wake up before the crack of dawn and punch in at 5:30 a.m. most days. It’s not easy work.
“The lumber is coming at you at a pretty good rate of speed,” Koosmann says. Once it reaches the grading station, he has just seconds to flip the board, scan for defects, determine which of four categories it belongs to and give it the mark that will ultimately determine its value. Over the course of nine hours, Koosmann might examine upwards of 3,000 ten-foot-long boards. That adds up to anywhere from 32,000 to 39,000 pounds to lift and turn, equivalent to one and a half school buses sans children.
The average mill worker in Wisconsin makes around $13.50 per hour. Graders earn slightly more — about $14.75, according to Zip Recruiter. That’s well over Wisconsin’s dismal $7.25 minimum wage, but not much more than what less dangerous jobs, as in retail, pay.
Grading lumber is also an art form. Since a grader’s mark determines the board’s price point, accuracy is crucial. Per NHLA standards, graders are required to complete an exhaustive eight-week course in Memphis, Tennessee to learn the knots and boules of hardwood. The full curriculum costs anywhere from $500-$2,600, a cost that’s often incurred by employers. Even after graduation it takes several years of real-world experience to become a grading master, an investment that fewer and fewer people are willing to make.
When Delegan graduated from NHLA grading school in 1996, there were sixty-four students in his class. The most recent class graduated five. The decline isn’t just a product of the pandemic; July 2019 saw seven graduates, and by all accounts attendance has been trending downward for years.
In contrast to its human counterparts, wood grading AI uses a laser to scan each piece of wood like a barcode as it zooms down the line, constructing an image pixel by pixel. Then, drawing on a database of hundreds of thousands of pictures, the machine’s algorithm compares the new board against similar specimens. From there it assigns a grade with (in the words of one manufacturer) “unparalleled” accuracy. Theoretically, machine precision reduces the number of cuts needed for each piece of wood. Fewer cuts means more yield and longer-lasting saws, which also means more revenue.
In October 2019, North Country Lumber finally bit the bullet, sending representatives down to Kentucky to watch an AI wood grading demonstration by the Italian logging tech firm, Microtec. Koosmann, whose job is listed as 100% automatable by Marketplace, was not among the representatives chosen to observe the AI in action.
Microtec currently supplies AI wood grading technology to over two hundred companies, including U.S. giant Idaho Forest Group. Their recent acquisition of Oregon-based rival, Lucidyne, promises to expand that number even further.
“That’s going to be a huge difference for small sawmills, middle sawmills, and large sawmills,” says Dercilio Lopes,an Assistant Research Professor who develops AI for lumber grading at Mississippi State University.
However, according to Lopes, the current AI needs improvement. When the algorithm identifies a knot in the wood, it draws a digital “box” around it — a little square of pixels. But, Lopes says, “knots, they’re not square. So basically, you lose area.”
Gazo, who worked with Microtec to develop their grading technology, doesn’t think that artificial intelligence can currently match the human eye. “If you look at how many cells and receptors are in your eye, and the connection between the eye and the brain,” he says, “even the best supercomputer today still cannot match it.”
But AI is trendy, he says, and neural networks don’t tire out or take sick days. The resulting “tech push” has created a sense of urgency around AI upgrades, which is great for companies like Microtec.
For now, North Country Lumber has refrained from investing in the technology — electing to pass on an upgrade for the reasons Lopes cited. “The artificial intelligence isn’t quite there yet,” says Delegan. Microtec declined to comment on their product’s efficacy.
If and when the mill eventually upgrades to AI, Koosmann won’t be primarily grading wood any longer. Instead, he’ll take over the machine’s technical operations and serve as backup grader when needed.
“Right now, it’s been a struggle for everybody… There’s not a lot of big paying jobs” he says. If you find a job you like and you can change with the times, “you hang onto it.”
Joanna Thompson is a science writer hailing from Tennessee, now based in New York City. You can sometimes find her online @jojofoshosho0.