Can AI fix what cheap labour cannot in Bangladesh? - The Daily Star

TEXT START: Bangladesh's garment industry is often explained by its low wages, yet research conducted inside its factories shows that wages do not explain why some firms perform far better than others.

The Dissection

This is a competent managerial diagnosis wrapped around a structurally evasive question. It correctly demolishes the claim that cheap labour alone explains factory performance, then relocates competitiveness to supervision, data, infrastructure, energy, finance and process discipline.

Its real question is not whether garment workers retain productive participation. It is which factories survive, which capture orders and which convert AI into measurable output. The article acknowledges displacement, gender exclusion, worker surveillance and buyer power, but treats them as implementation problems. It keeps ownership and control of AI capital outside the frame.

The distinction between automation and AI is useful at the operational level. It is not a systemic defence. Sewing's resistance to full automation is a lag condition, not a permanent human moat. The article measures the delay while implying that the delayed endpoint remains negotiable.

The Core Fallacy

The central error is treating AI as a productivity tool whose value depends mainly on management quality, rather than as capital that progressively absorbs management, coordination, inspection, planning and operational judgment.

Better management does not defeat the Discontinuity Thesis. It prepares the factory to implement it. The firms with cleaner data, stronger supervisors and tighter processes will be the first to deploy AI, reduce labour requirements, standardise decisions and widen their advantage. Weaker firms may catch up technologically and still destroy jobs in the process.

The article asks whether AI will narrow or widen productivity gaps. That is a firm-selection question, not a survival question for the wage-consumption circuit. Under P1, cognitive work becomes cheaper to automate. Under P2, institutions cannot preserve stable human-only domains at scale. Under P3, productive participation collapses. A factory can remain globally competitive while its workforce becomes economically redundant.

The article also confuses industrial scale with social viability. Bangladesh may retain export volume, but the gains can flow to owners, buyers and software vendors while workers lose bargaining power, wages and control over the data generated by their labour. Output can survive the workers who once made it socially useful.

Hidden Assumptions

• Management capability is treated as a durable human moat rather than a capability AI can codify and replicate.
• Productivity gains are implicitly expected to support workers or suppliers, despite no mechanism ensuring that distribution.
• Labour displaced by automation will be replaced by higher production volumes, new roles or rising wages.
• Sewing's physical difficulty will remain a meaningful defence long enough to prevent eventual substitution.
• Energy and finance shortages will merely delay AI adoption rather than intensify selection in favour of firms already closest to the frontier.
• Measurement, pilot programmes and stop dates can control a competitive race in which cost reduction is the governing pressure.
• Training women into supervisory roles solves a durable power structure rather than temporarily increasing their value before supervisory work is also automated.
• Buyer demand for traceability will create supplier bargaining power instead of giving buyers more surveillance and leverage over factories.
• A three-to-five-year research horizon can meaningfully answer a structural question whose endpoint is institutional and generational.

Social Function

This is transition management with a partial-truth and ideological-anesthetic function. It is not crude copium; it openly records layoffs, weak infrastructure, buyer capture and gendered exclusion. That makes the anesthesia more effective.

The article converts a power crisis into a sequencing problem: measure first, train supervisors, run AI trials, fix energy, improve finance. Those recommendations may help selected firms navigate the transition. They do not answer who owns the machines, who controls the data, who receives the surplus or what happens when most workers are no longer economically necessary.

Its implicit message is that Bangladesh can remain viable if its factories become competent enough. The harsher reality is that competence may determine which firms administer the carcass, not whether the carcass remains a mass-employment system.

The Verdict

The article is accurate about proximate causes and inadequate about systemic consequences. Cheap labour does not explain why factories differ; organisational capability does. But organisational capability is not an escape from AI-driven obsolescence. It is the launch infrastructure for it.

The likely DT trajectory is clear: better-managed factories adopt AI and automation faster, win the remaining orders, require fewer workers and force weaker firms out. Energy shortages, expensive credit, gender barriers and trade-policy shocks can delay or redirect the process. They cannot restore the wage-to-consumption circuit.

This article diagnoses how Bangladesh's garment industry may be sorted inside the collapse. It does not demonstrate that the industry can preserve mass productive participation. Under the Discontinuity Thesis, the factories may survive. The labour system does not.