Ten Years After Rana Plaza: Where Are 5 Million Workers Heading with AI Algorithms?

On the outskirts of Dhaka, Bangladesh, the owner of a medium-sized garment factory supplying European fast-fashion brands stared blankly at the order contract in his hand. Not because there were no orders—buyers were still placing them. But because the contract now included an extra line: "Suppliers shall complete a production line automation transformation plan and submit a worker skill transition training program by 2027."

He flipped the contract to the payment terms page. The unit price was the same as last year.

Looking back a few months, in December 2023, the Bangladeshi government raised the minimum wage for textile workers by more than 50%, to about $95 per month (according to Bangladesh's Daily Star). International brands publicly called this a "milestone progress."

But this factory owner did the math: The factory currently employs 1,200 workers, with a total monthly wage bill of approximately $114,000. The upfront cost of an automated cutting system plus an intelligent overhead conveyor system is about $1.8 million in the first year, but thereafter only maintenance and electricity costs remain—no wage increases, no social insurance payments, and certainly no dealing with strikes.

Which option would he choose?

This is not a multiple-choice question unique to Bangladesh. It is a systematic redistribution of costs happening across the global apparel supply chain. And pushed to the center of the chessboard are those 5 million workers operating sewing machines in 40°C workshops.

1. Two Numbers from a Pile of Rubble

Let us first return to April 24, 2013, in the Savar district of Dhaka, at Rana Plaza.

This eight-story commercial building collapsed during the morning rush hour, burying five garment factories that were working inside. Final death toll: at least 1,132 people. Injured: more than 2,500 (according to a 2023 investigation report by Tricontinental: Institute for Social Research).

The day before the collapse, authorities had issued an evacuation warning. But the building owner, Sohel Rana, told factory owners: "The building is only slightly damaged." Factory owners told workers: "Keep working—rush this order." The supplier list included brands such as Benetton, Zara, Walmart, Prada, and Gucci.

Eleven years have passed. The rubble of Rana Plaza has long been cleared, and Bangladesh's garment industry seems to have emerged from the shadow. It is now the world's second-largest apparel exporter, with textiles and garments accounting for over 80% of the country's total exports, reaching approximately $48 billion in the 2023–2024 fiscal year (according to data from Bangladesh's Daily Star and Xinhua Finance). There are more than 4,000 garment factories nationwide, providing over 5 million jobs. International labor rights organizations, brands, and the Bangladeshi government have jointly launched a factory safety inspection system billed as the world's strictest.

But tucked away in the Tricontinental report is a number almost no one noticed: In the decade following the Rana Plaza collapse, at least 109 additional buildings collapsed in the same region of Bangladesh, killing at least 27 workers.

109 buildings. This is not a failure of the safety system—this is a performance of the safety system.

And the second number I want you to remember is 10.97%.

According to data from the U.S. Department of Commerce's Office of Textiles and Apparel, in the first half of 2024, Bangladesh's apparel exports to the United States fell by 10.97% year-on-year (as cited by Xinhua Finance). During the same period, Vietnam's apparel exports to the U.S. surged by about 17% (January–July 2025 data from the Vietnam Textile and Apparel Association). In 2025, Vietnam briefly surpassed China to become the largest apparel supplier to the United States.

Put these two numbers together, and a clear causal chain emerges—not that factories became more dangerous after Rana Plaza, but that global buyers assessed the comprehensive risks of Bangladesh as a "production workshop" (including workers' ability to demand rights, the government's attitude toward labor organizations, the stability of energy supply, and port clearance efficiency) and made a business decision: shift incremental orders to production bases with lower risk.

Here is a highly counterintuitive truth: Brands are not afraid that factories are unsafe; they are afraid of the chain reaction caused by unsafe factories being exposed—worker strikes, order delays, social media crises. The risk is not "danger to life"; the risk is "disruption of delivery."

2. $95 a Month—A "Conscience Account" That Pushes Workers into the Algorithm Black Hole

In December 2023, the minimum wage for Bangladeshi textile workers was raised by more than 50%. Sounds like good news. In press releases from international media, NGOs, and brands, this was a victory.

Let's break down this account.

A monthly salary of about $95. Given the current cost of living in Bangladesh, this is barely subsistence level. But from the factory owner's perspective, this is an unstoppable increase in the cost structure—the garment processing industry survives on unit profits of just a few cents.

The Tricontinental investigation recorded a factory owner's words: "Factory owners pursue profit maximization... they will not pay overtime... partly it can be blamed on brand retailers... Even if the profit is only 2-3 cents, it can make a big difference, but these companies do not want to include compliance in their cost calculations."

This statement carries a lot of weight. It admits two things: first, factory owners are exploiting workers; second, brands are complicit in this exploitation because if prices rise, they move orders. The entire chain is a tight string—any excessive force on either end will snap it.

Now, the minimum wage string has been pulled up by 50%. The factory has only three paths:

1. Raise prices—Brands won't allow it because Vietnam is waiting to take orders.
2. Lay off workers—Replace "optimizable" workers with AI scheduling, intelligent overhead conveyors, and automatic cutting machines (one automated cutting unit can replace 8-10 workers—an already mature industry solution).
3. Close down—If the factory is too small to afford equipment, it simply goes bankrupt.

All three paths lead to the same outcome: Workers are pushed out of the production line.

Therefore, the current wave of garment factory closures in Bangladesh did not happen overnight. It is the result of a chain reaction: brands demand a "just transition" → minimum wage rises → factory costs skyrocket → brands refuse to bear the cost → factories invest in automation to replace labor → low-skilled workers lose jobs. This is not a "natural disaster"; it is a path of interest transfer meticulously calculated by global buyers.

A research report published by Amnesty International in November 2025 put it even more bluntly. Secretary General Agnès Callamard used a phrase rarely seen in NGO discourse, calling it "a 'sinister alliance' between fashion brands, factory owners, and the governments of Bangladesh, India, Pakistan, and Sri Lanka."

The study was based on 88 interviews at 20 factories in four countries (September 2023 to August 2024). They sent questionnaires to 21 major brands. Only Adidas, ASOS, Fast Retailing, Inditex, Otto Group, and Primark provided complete responses. Boohoo, H&M, Desigual, Next, and Gap provided no information at all.

When a country accounting for 80% of global apparel exports pays its workers just $95 a month, and brands respond by "not providing information," this is no longer a discussion about "labor rights"—it is using silence to maintain the legitimacy of structural violence.

3. "AI Replacement" Is a Carefully Crafted Narrative

Now we get to the core of this article—the thin layer of paper that needs to be pierced: Almost all the reports you read about "AI replacing textile workers" may have been written from the buyer's perspective.

Because this narrative has a fatal logical flaw: It assumes AI replacement is a natural result of technological progress, while completely ignoring who chose to press the "start automation" button and who set the economic tipping point for replacing humans with machines.

Let's do a thought experiment. Suppose:

• Brands are willing to pay an extra $0.50 for each T-shirt produced in Bangladesh;
• This $0.50 is enough to cover the minimum wage increase and leave a surplus for improving working conditions;
• Then, would the factory owner still have the incentive to take out a $1.8 million loan to buy automation equipment?

The answer is: Maybe, but not under compulsion. They could choose automation to improve efficiency, but not because they would die without it.

But in reality, brands did not pay that extra $0.50. They found alternatives in Vietnam that are equally cheap but have more stable supply chains, friendlier tariffs (Vietnam has signed 17 FTAs), and lower political risk. They pressed the "transfer order" button and told Bangladeshi factories: "Either find your own way to improve efficiency, or we leave."

Thus, we see: The Daily Star reported that one study predicts the online clothing market in the U.S., EU, and Africa will reach $308 billion by 2026. If cross-border e-commerce platforms are built, Bangladesh could earn $489 million in export revenue by 2027. That number was also mentioned—accounting for 0.2% of the U.S. online market, 0.1% of the EU market, and 0.75% of the African market.

Of that $308 billion global pie, Bangladesh can only scrape off a crumb.

This is not AI eliminating workers. This is the global profit distribution system "eliminating" the bargaining power of Bangladesh's textile workers as a whole.

The World Economic Forum predicts that by 2030, 22% of global jobs will be transformed, with 92 million jobs displaced and 170 million new ones created, resulting in a net increase of 78 million jobs (according to the WEF's Future of Jobs Report, cited by Qiushi Journal). But such predictions share a common blind spot: They assume the new jobs and the displaced jobs occur in the same country and among the same people.

In reality, this is not the case. The displaced worker is a 23-year-old female sewing operator in Bangladesh; the newly created job might be an AI trainer in Silicon Valley. The "net increase" in global employment opportunities masks a brutal mismatch across geographical and class dimensions.

4. Energy Loans and Port Reforms—The "Lubricant" Accelerating Worker Elimination

Here I want to make an even more counterintuitive judgment: Japan's assistance to improve Bangladesh's energy supply and the Bangladeshi government's push to reform Chittagong port clearance—both seemingly "good things" that enhance Bangladesh's competitiveness—may, in the context of AI replacing workers, actually serve as "lubricant" accelerating the elimination of traditional workers.

First, consider energy. Bangladesh has long suffered from unstable electricity supply. Garment factories rely on diesel generators to survive, at exorbitant costs. The energy loan program from the Japan International Cooperation Agency (JICA) could theoretically stabilize the grid and reduce factory electricity costs.

The problem lies with the word "stable."

What does more stable electricity mean? It means automation equipment—automatic cutting lines, intelligent overhead conveyor systems, AI vision inspection machines—can run 24 hours a day without interruption. The logic is simple: In an unstable power environment, even the most advanced machinery could be idle for half a day due to a power outage. Once electricity is stable, the payback period for automation investments shortens dramatically, and the investment appeal rises sharply.

Now look at the port. Chittagong Port handles about 92% of Bangladesh's total import-export trade and 98% of containerized cargo. In 2025, it handled 3.409 million twenty-foot equivalent units (TEUs) (according to Chittagong Port Authority data). Improved clearance efficiency and reduced logistics costs are indeed beneficial for textile exports.

But "faster logistics" also redefines workers' circumstances.

The competition among fast-fashion brands is essentially a competition for speed. ZARA can compress the time from design to store shelf to 15 days. Whoever can ship goods faster, clear customs faster, and deliver to warehouses faster gets more repeat orders. And this "faster" demands from factories extremely high production flexibility and management precision—quick style changes, precise production scheduling, and digitalized quality control. These capabilities are precisely the weakest points of traditional operators with relatively lower education levels.

When Chittagong port was terribly congested, global buyers positioned Bangladesh as "cheap, don't expect on-time delivery, but cheap enough to tolerate delays." When the port becomes efficient and logistics speed up, buyers immediately upgrade their requirements to "cheap, on-time, and capable of small-batch quick response." The ultimate solution to these demands is not making existing workers more disciplined, but using systems to replace human decision-making, cameras to replace human monitoring, and algorithms to assign the sewing sequence of each garment.

Thus, energy loans and port reforms are not "helping" Bangladeshi workers stay in their jobs—they are giving capital more confidence to deploy automation here, and automation replaces whom?

Still those workers who, when energy was unstable and the port was congested, used sweat and excessive overtime to rush orders.

5. India's PLI Scheme—"Cheap" Is No Longer a Barrier a Single Country Can Monopolize

Zooming out, you will find that Bangladesh's predicament is not isolated. The entire South Asian textile triangle—China, India, Bangladesh—is undergoing a drastic positional reshuffle.

India's PLI (Production Linked Incentive) scheme for textiles has entered its third execution phase, with approved company lists and investment amounts finalized. The scheme provides particularly clear incentives for investment in MMF (Man-Made Fibers) fabric manufacturing, with substantial subsidies directed toward expanding polyester, nylon, and viscose production lines (according to India's Ministry of Textiles PLI scheme announcement).

This is a variable that has received relatively little discussion in domestic industry circles but has far-reaching implications. In the past, the rough division of labor in global textiles was: China strong in man-made fibers, India strong in cotton spinning, Bangladesh strong in cotton garment processing. But this division is being disrupted by India's PLI scheme.

When India's domestic MMF fabric capacity increases, Indian textile exports to Europe and the U.S. will no longer be limited by high tariffs on cotton products and seasonal fluctuations (cotton prices are heavily affected by climate; polyester by crude oil). India can supply MMF knitted fabrics and apparel to the global market year-round at prices close to Bangladesh's. Vietnam has already blazed this trail—leveraging FTA tariff advantages with the EU, Japan, and other markets to make MMF sportswear and knitwear part of the core supplier lists of global brands.

For Bangladesh, this competitive landscape is particularly treacherous. Because cotton products still account for a relatively high proportion of Bangladesh's garment exports, while global consumer purchasing preferences are shifting toward MMF sportswear, outdoor clothing, and stretch casual wear. This structural shift, combined with the time window of India's PLI capacity expansion, means that Bangladesh may, within the next five years, descend from being "an almost irreplaceable cheap manufacturing base" to "a backup capacity that is no longer the optimal option."

Chinese fabric companies are also adjusting their strategies. Facing competition from Indian MMF fabrics, Chinese companies still have a clear first-mover advantage in high-end differentiated product lines—functional polyester (moisture-wicking, cooling, far-infrared), micro-denier knitted fabrics, three-proof finishing, and other niche categories where technology and quick response capabilities are beyond India's short-term reach. But the share of low-end conventional fabrics is being rapidly eroded, and this is precisely the raw material structure that Bangladeshi garment factories rely on most.

Bangladesh is squeezed from both sides: above, China's high-end fabric branding; below, India's MMF capacity expansion and Vietnam's tariff channels. It is trapped in the middle, and its only selling point—"ultra-cheap"—has now lost its marginal advantage after the $95 base rose by 50%.

6. BGMEA's Support for Small Workshops—Extending Life or Delaying Execution?

The Bangladesh Garment Manufacturers and Exporters Association (BGMEA) has in recent years been promoting a strategy: subcontracting part of large factories' orders to compliant small workshops, while providing technical training and management support to those workshops.

There are two diametrically opposed interpretations of this logic within the industry.

The optimistic camp believes that this can ease the automation pressure on large factories—large factories use AI systems to manage order allocation and production scheduling, while small workshops handle flexible sewing tasks, similar to the division of labor in China's early garment industry clusters. Large-scale factories retain core high-value-added processes (cutting, finishing, inspection) and subcontract simple sewing operations to community-based workshops, thereby maintaining more jobs.

But contacts I have on the ground in Bangladesh's supply chain offer a colder assessment: This model is essentially an apology for "casualization."

Why? Because small workshops cannot afford the cost of safety upgrades—LEED certification, fire safety retrofitting, wastewater treatment—any single item could wipe out more than half a year's profit for a small workshop. When orders flow from formal factories to small workshops, workers' social insurance, contracts, and minimum wage protections are all eroded by the shell of "flexible employment." The difficulties mentioned in the Amnesty International report about workers in Bangladesh's special economic zones being unable to form unions are magnified tenfold at the small workshop level: there is no union at all, and there may not even be a formal employment relationship.

At the same time, this subcontracting model provides a convenient "firewall" for brands: If an accident occurs in a small workshop (fire, collapse), the brand can declare, "We signed compliance clauses with our tier-1 suppliers and were unaware of their subcontracting practices." In fact, such declarations have appeared in every fire investigation in the past.

Therefore, BGMEA's support for small workshops, on the surface, is about "protecting jobs," but the deep logic is to use subcontracting channels to "invisibly downgrade" the labor costs of formal factories, buying time for large factories to automate. This has nothing to do with the "just transition" talk of skills training and job conversion.

7. AmCham and Investment Reforms—Who Is Investing in Bangladesh? Money Comes, People Leave

The American Chamber of Commerce (AmCham) has been actively pushing Bangladesh to improve its investment climate, including simplifying approval processes, relaxing foreign direct investment (FDI) restrictions, and establishing special economic zones.

Surface logic: Attract foreign capital to build new factories and create jobs.

What actually happens? Foreign capital enters through this path:

1. Invest in building modern factories (automated cutting lines, AI quality inspection systems);
2. Introduce management software (ERP+MES systems integrated with brands' PLM systems);
3. Hire fewer workers than local factories, but with higher education and technical requirements;
4. Then "suck" orders away from surrounding local factories.

Thus, you see a bizarre pattern: AmCham urges Bangladesh to "improve the investment climate," foreign capital does come in, but local old factories begin to close because they cannot get enough orders. The few who are hired by the new factories are young technicians who can operate the systems; those who lose their jobs are middle-aged female workers who have been operating sewing machines for 15 years in the old factories.

This is not capital creating employment; this is capital redistributing who is qualified to survive in the job market.

Let's pull back another data point: carbon emissions. According to industry public estimates, Bangladesh's garment industry needs to reduce carbon emissions by 50% by 2030, requiring approximately $6.6 billion in investment, with a funding gap of about $4.8 billion (according to industry reports cited in Wind Review). $2 billion has already been invested in environmental compliance and sustainable practices (including the LEED Platinum-certified Hams Garments Limited factory, which scored 108 out of 110, the highest score globally for a garment factory).

Key question: Whose job opportunities did these environmental investments and foreign capital modernization investments convert into?

The answer is not "textile workers," but "sustainability compliance managers" and "ESG reporting specialists."

8. Those 5 Million Workers—Who Speaks for Them?

After all this data, policy, and supply chain structure analysis, this final section has no more data. There is something harder.

In the 88 interviews conducted by Amnesty International, more than two-thirds of the respondents were female workers. In Bangladeshi textile factories, female workers form the vast majority—their hands are said to be more dexterous, more compliant with management, and less likely to form unions. When a factory owner explained why violence was used to suppress a worker protest, the same factory owner told the researcher in a calm tone: "Partly it can be blamed on brand retailers... Even if the profit is only 2-3 cents, it can make a big difference."

The three layers in this statement represent the entire moral spectrum of the global apparel supply chain:

• Workers earn $95 a month;
• Factory owners survive on 2-3 cent profit margins;
• Brands use silence to write maximum profit into contracts.

Now, a fourth layer is being written into this system: AI production scheduling software.

When a female sewing operator's hands become a "unit cost variable" in a machine learning model, she is no longer a "vulnerable group needing protection"; she becomes a parameter that can be optimized out in real time.

In May 2025, China's Qiushi Journal cited ILO data: One-quarter of global jobs may be impacted by generative AI. But ILO senior economist Janine Berg added a sentence: "We can easily get caught up in the hype around artificial intelligence. What we need is clarity and specific context."

What is the specific context?

The specific context is lunchtime at a factory in Dhaka, where a thousand female workers squat at the factory gate eating their meals. Three kilometers away, in a new building constructed after Rana Plaza collapsed, another factory owner is demonstrating to a German buyer via video conference: This production scheduling system can increase efficiency on each line by 23%. The German buyer says, "Very good." "So, can we reduce the unit price by another 3 cents?"

This industry's story is not about "whether workers can learn to operate AI." The real story of this industry is: After global brands locked the profit margin to "every body equals 2 cents of cost," they suddenly announce, "Now we want to upgrade from 2 cents to more efficient code"—where should those ten million feet that once pushed sewing machines go?

Eleven years ago, when Rana Plaza collapsed, the international community said it would never happen again. Then another 109 buildings fell. Now it is not buildings that are collapsing—it is the system collapsing: a world factory built on cheap lives, and its builders are pulling the first brick out from the bottom.

No one knows what will happen when the last brick is pulled. But the sound should have already started to ring.