Kiran Kumari sits behind her Brother sewing machine on a vast factory floor, dressed in a purple and gold sari and flanked by dozens of fellow clothes stitchers. With a deft sweep of her arm, she picks up a piece of cloth from the bundle next to her, places a white panel on top and feeds them both quickly under the needle. It is just one of about 400 collars Mrs Kumari will sew on to Ralph Lauren tops during her eight-hour shift, each one taking a few minutes to complete and all done for about $100 a month. 

She and the other 4,800 or so garment workers spread across three Matrix Clothing factories south of Delhi form just a small part of the army of inexpensive labour that many developing countries hope will propel them towards prosperity. This is particularly true in South Asia, where populations are growing fast and wages remain low. The World Bank estimates this region alone will add 1m to 1.2m new workers to the labour market each month for the next two decades — 240m people.

Yet 8,000 miles away in the US city of Atlanta, a robotics company is working on a machine that could in time put Ms Kumari out of work for good. The “Sewbot” technology, being developed by Softwear Automation, aims to automate the entire clothes-making process.

The technology is still years away from becoming cheap and reliable enough to replace humans. Ms Kumari, for instance, earns about $1,200 a year. The company will not say how much the Sewbot costs, but industry sources say it will run into hundreds of thousands of dollars. But with automation sweeping through established industries, experts are warning that it is only a matter of time before this technology undermines the economic model of large parts of the developing world.

South Asia is especially at risk, given how much of the region’s economic plans rely on mopping up the international manufacturing work for which China is becoming too expensive. In India, Pakistan and Bangladesh, policymakers talk about reaping a “ demographic dividend”, as populations grow rapidly while average wages remain about a quarter of those in China.


Yet economists are starting to ask how much of a dividend these young, cheap and potentially restless workforces will enjoy when robots are increasingly able to perform the kinds of labour-intensive manual work on which they rely. The impact is also likely to be felt in places such as Southeast Asia, another garment hub, and sub-Saharan Africa.

“Robotics and artificial intelligence are the next revolution,” says Rajiv Kumar, an economist and founder of the Pahle India Foundation. “They are going to be more disruptive than any of the revolutions we have seen in the past — steam, electricity, the assembly line or computers — because they are going to replace not just routine but complex mental functions. The fear is that our so-called demographic dividend could become a demographic nightmare.”

***

Several years ago academics noticed that something strange had been happening in parts of Asia, Latin America and sub-Saharan Africa. While many countries had been growing strongly, the share of jobs in the manufacturing industry had barely risen since the 1980s, and in some cases had begun to fall — long before economists expected it to happen.

A robot sewing process created by US start-up Sewbo which will allow computers to control the stitching of T-shirts © Company

In 2015 Dani Rodrik, a Harvard economist, coined the phrase “premature deindustrialisation”, arguing that many developing countries were becoming service-based economies much earlier in their histories than their western equivalents. Technological change had a large part to play, he said, warning that the trend could have grave consequences for economic growth and political stability in those regions. “Manufacturing has traditionally absorbed significant quantities of unskilled labour,” he wrote, warning that the deindustrialisation trend was “not necessarily friendly to liberal democracy”. 

His findings helped explain a study by the International Labour Organisation and the UN Development Programme soon after the financial crash. That report found that from 2003 to 2009, jobs growth in South Asia was just a third of the level of overall economic growth. Experts warned the region was experiencing “jobless growth”. 

Since then, the pace of technological change has increased and some industries are losing ground altogether. India’s IT services companies, which have flourished over the past three decades, have begun to lose ground to automatic cloud-based computing systems. Two of the sector’s biggest companies, Infosys and Tata Consultancy Services, have both shed jobs this year. Meanwhile in Chennai, 400 robots have replaced humans in large parts of the Hyundai car plant. 

A graphic with no description

The chief executive of one big Indian technology company, who asks not to be named, says the drop in jobs would be worse if bosses were not concerned about the consequences from laying off so many workers. “We carried out an audit and found that we could replace half our staff with artificial intelligence,” he says. “We would do it, but for the social fallout it would create.”

Chief among these is the clothing industry. Bangladesh, in particular, has become so reliant on its garment sector that 82 per cent of all its exports are clothes, and 2.5 per cent of the entire population has a job making them. Across Bangladesh, India and Pakistan, there are about 27m people employed by the industry, according to the Clean Clothes Campaign, which lobbies for better conditions for those workers. 

One reason this industry is generating jobs is because South Asian workers are undercutting their counterparts in China. Average hourly wages for Chinese factory workers hit $3.60 last year, up nearly four times from a decade ago, according to market researchers Euromonitor. The average Chinese factory worker now earns about five times as much as their counterpart in India, and is paid close to those in Portugal or South Africa. 

Software Automation Vice President Sales and Marketing Pete Santora stands next to a traditional sewing machine while holding up a camera which is used in the automated robotic machines they call "sewbots", pictured Wednesday, June 28, 2017, at the company's facility in Atlanta. The camera in a sense becomes the eyes of the seamstress and guides the robotics during the new process of making clothing. (John Amis)
Softwear Automation’s Pete Santora with the camera used in the Atlanta company’s Sewbots © John Amis/FT

The other reason is that the industry has proved curiously impervious to automation. Making a T-shirt is almost exactly the same process today as it has been since the invention of the automatic sewing machine in the 19th century. T-shirts are usually made from cotton, a flimsy material that bends and curls, making it difficult for a robot to pick up and move precisely. Behind Ms Kumari’s quick and almost casual stitching action lies a host of tiny processes and decisions so complex that the most sophisticated software engineers struggle to replicate it.

“There are four processes that go into making an item of clothing,” says Gautam Nair, managing director of Matrix Clothing. “There is picking up the item, aligning it, sewing it and disposing of it. Of these, only the sewing has so far been automated, and the sewing machine came in a long time ago. The other parts of the process are still done more quickly and more cheaply by humans.”

Workers at a garments factory in Hanoi, Vietnam

***

This may be about to change. In a former cabinet-making factory in Atlanta, a group of Softwear Automation technicians are huddled around touchscreens punching in computer code as they try to perfect the Sewbot. To address a problem with delicate materials they have given the robot cameras, which function as eyes, just as engineers have done with self-driving cars. The cameras take shots of the material being stitched, analyse them and then guide the movements of the robotic arms. 

The technology has caught the attention of Walmart, the world’s largest retailer, which has given the company $2m as part of a project to automate the production of jeans. In September, the Sewbot made a breakthrough, successfully sewing an outside seam on to one pair. Next year, the company plans to expand to T-shirts, 97 per cent of which are produced outside the US.

Palaniswamy Rajan, Softwear Automation’s chief executive, explains the commercial imperative behind developing technology of this kind, even when clothing companies have a vast army of cheap labour at their disposal. 

“If you’re Macy’s [the department store], and you want to get 100,000 pieces of this style, you order from China nine to 12 months ahead of time,” he says. “If you produce closer to the consumer, you could buy an order of 10,000, get it in a month, and see if certain designs sell better than others.” 

Nate Damen of Softwear Automation attaches an end-effector to the robot arm of one of the company's machines © John Amis/FT

Walmart is not the only big retailer looking at clothing automation. Amazon filed a patent in April to develop a “stitch on demand” machine that would automatically make clothes after they have been ordered. Yet US companies are being spurred on not just by the pace of technological change but by political reality. The election of Donald Trump as president and the promise of “America first” trade policies have led companies to start looking at ways they can bring jobs back to the US. 

“No doubt our work will be different in the future,” Doug McMillon, Walmart chief executive, told staff recently. “Robots, drones and algorithms will do some work that we used to have to do. Some people are afraid of what these changes will bring. I don’t think we should be . . . The secret to our success will always be our people.”

Policymakers in developing countries are aware of these trends. Some believe there is still time to generate a South Asian manufacturing boom as long as governments get the policies right. 

Arvind Subramanian, chief economic adviser to Indian prime minister Narendra Modi, has studied premature deindustrialisation. Sitting in a large office at the heart of New Delhi’s finance ministry, he says: “Yes, robots have started cutting soft cloth. But I don’t know if it is the nine- or 10-year horizon we should worry about, or more realistically the 20-year horizon.”

Mr Subramanian says a mixture of good training, carefully structured free trade agreements and labour reforms could all generate manufacturing jobs before robots get sophisticated and cheap enough to replace them. If he is wrong, the social impact would be huge, warns Mr Kumar of the Pahle India Foundation: “It could be a nightmare because of all the educated, aspiring young people who would be unemployed as a result of this automation.”

Others argue that countries should cut their losses and focus on service industries. Uri Dadush, a senior fellow at Morocco’s OCP Policy Centre think-tank, wrote in 2015 that “the importance of manufacturing as a development lever is declining”. He argues that countries such as India, Pakistan and Bangladesh should focus on turning themselves into specialised hubs for certain sectors, such as tourism, transport or finance. 

For Mrs Kumari, the prospect of a jobs boom in services to replace those lost in manufacturing is of no consolation.

“If I wasn’t working in garments there is nothing I could do,” she says. “There is nothing else I am trained to do.”

Letter in response to this article:

A term used by Unctad economists since 2003 / From Richard Kozul-Wright, Unctad

Get alerts on AI and jobs when a new story is published

Copyright The Financial Times Limited 2019. All rights reserved.
Reuse this content (opens in new window)

Follow the topics in this article