There’s one part of fashion that the technological leaps of recent decades still haven’t managed to revolutionise: sewing.

Even as complex products like cars and computers involve more robots and automation in their manufacturing, clothing continues to be made by humans manually guiding fabrics and operating sewing machines to assemble finished garments. Many of these workers are poorly treated and poorly paid.

It’s not for lack of trying that fashion hasn’t managed to automate sewing. In 2016, a Seattle start-up called Sewbo claimed it produced the first robotically sewn garment, a T-shirt. At the same time, another company, SoftWear Automation, was emerging as a leader in the field with a line of sewing robots it developed. In 2017, it partnered with Chinese manufacturer Tianyuan Garments to build an automated production line for T-shirts in an Arkansas factory it planned to open in 2018. Then, supply-chain giant Li & Fung announced a partnership with the company “to accelerate the full digitisation and automation of the manufacturing process.” Robots seemed poised to finally take over the job of sewing.

But since then… nothing. What happened?

“We ran into technology challenges working with knitted fabrics,” said Palaniswamy “Raj” Rajan, chief executive of SoftWear Automation.

Fabric is difficult for robots to work with, because unlike, say, sheet metal, it bends and stretches. Handling it requires constant minute adjustments that are natural for humans but much harder for machines. The construction of knit fabrics, which gives them their stretch, makes them even trickier than wovens. Rajan described T-shirts as “a simple garment but a complex fabric” since there’s so much variability from one material to the next.

SoftWear Automation’s T-shirt factory still isn’t operational. It plans to open next year in a location still to be determined. It is no longer working with Li & Fung, Rajan said, and will not sell or lease its robots. Its focus will be on premium, on-demand, small-batch production runs, most likely for independent brands, as it works to scale and expand into items other than T-shirts.

“We anticipated two years and it’s taken us five,” Rajan said, chalking it up to the unexpected delays that often come with developing new technology.

At some point, technology likely will change how clothing is made. It’s happened before. “What the telegraph is to the commercial world, the reaper to the agricultural, the sewing-machine is to the domestic,” the New York Times declared in 1860, ranking the sewing machine among the “most important of the labour-saving inventions.”

But it’s unclear how soon that might occur. There’s little urgency to invest in the research and development needed to build better robots when fashion can continue to draw on a vast pool of inexpensive labour around the world to produce clothing at remarkably low costs. This paradigm has allowed the industry to sell more clothing at lower prices, contributing to the environmental destruction fashion is now trying to mitigate. Though as wages rise in countries like China and pressure to improve labour conditions in garment-making nations like Bangladesh grows, it could raise costs enough to make automation look increasingly attractive.

But even for those willing to spend the money, automating production is a challenge. In 2017, I was among the first group of journalists to visit Adidas’s robotic “Speedfactory” in the small town of Ansbach, not far from Adidas’ headquarters in Germany. The team at Adidas walked us through the different processes, most of which were automated. It had partnered with Oechsler, a maker of parts for automotive and industrial companies, to build the robots.

At the time, it really seemed as if it could introduce a whole new paradigm in sneaker manufacturing. Rather than relying on a sprawling network of specialised suppliers around Asia to produce different components, Adidas had managed to consolidate everything into a single factory where robots did much of the work. This model was supposed to let Adidas respond faster to changes in the market and rely less on making large volumes of sneakers ahead of time.

But in 2019, the company announced it would cease production at the German Speedfactory and another it had built in the US, saying it would incorporate the technology at some of its Asian suppliers.

As capable as the robotic factory was, one thing it couldn’t do was make a wide variety of sneakers. It was ideal for producing certain styles of sneakers with woven uppers, but it couldn’t make shoes like the Stan Smith or Superstar, two of Adidas’ all-time money makers, which still required stitching together pieces of leather or other materials. Different robots would be needed for that job.

Adidas never expected full automation, though. Kasper Rorsted, the company’s chief executive, said in 2017 that goal wouldn’t be possible in five or even 10 years. There are some jobs robots still just can’t handle.

“The biggest challenge the shoe industry has is how do you create a robot that puts the lace into the shoe,” he said. “I’m not kidding. That’s a complete manual process today. There is no technology for that.”