When muscle stops paying

At the World Humanoid Robot Games in Beijing last month, a machine ran the hundred metres. It crossed the line in a little over 21 seconds. A fit schoolboy in Dhaka would beat that without training. In the relay, one robot collided with another, and both went down, and the crowd laughed the way crowds do at a sack race.

The laughing is fair. What follows is not comfortable.

Tap here to add The Daily Star as a trusted source

These games were held for the first time only last August, with 280 teams and around 500 robots. This year there were 666 teams and more than 2,000. The field roughly quadrupled in 12 months. The machines that fall over this year are the machines that will finish next year. Usain Bolt ran the hundred in 9.58 seconds. The gap looks enormous. It is not, really, because robots are improving in a year by more than humans have in a century.

We have watched this curve before. Not long ago, a chatbot could barely finish a coherent sentence. Today it writes contracts and code. The robot sprawled on a Beijing track is not a failure. It is a prototype.

About 1.13 million Bangladeshis went abroad for work in 2025, but roughly three in four were semi-skilled or unskilled, leaving them exposed to automation.

The jobs that go first are easy enough to predict. They are repetitive, physical and done in controlled spaces. Warehouse picking. Packing. Loading. Cleaning large standardised buildings. Basic assembly. Amazon already runs hundreds of thousands of robots in its facilities. China installs more industrial robots every year than the rest of the world combined. None of this is a forecast. It is already procurement.

Now hold that list against our own labour export. About 1.13 million Bangladeshis went abroad for work in 2025. Four percent were classified as professionals. Twenty-two percent were skilled. The rest, roughly three in every four, were semi-skilled or unskilled. Two-thirds went to a single country. Most went into construction, cleaning and domestic service, straight into the column that machines are being built to empty.

An average Bangladeshi migrant sends home around $200 a month. A Filipino sends more than double. The difference is not effort. Our workers are not outworked by anybody. This is the part policymakers need to sit with. An unskilled population will soon not be manpower. Just population. Manpower means something an employer abroad is willing to pay a premium for. Without that, our workers compete not only with Nepalis and Indians but with hardware that does not sleep, need a visa or send money home.

So what do we teach, and how fast? Start with English. It is the cheapest and quickest thing we can fix, and it explains much of the gap with the Philippines. English turns a cleaner into a supervisor and a helper into a receptionist. It also lets a worker read her own contract and argue her own case. We treat English as a subject for good schools. It should be treated as migration infrastructure.

Then basic IT. Not coding. Ordinary digital competence. Operating a laptop, using email and spreadsheets, filling online forms, handling scheduling apps, inventory systems and digital payments. A worker who can log a job on a tablet has a future in a workplace where one who cannot has already been replaced. Then aim at the work machines will reach last. Elderly care. Nursing. Physiotherapy. Wiring and plumbing in old, irregular buildings. Hospitality that depends on warmth. And technicians who install and repair the robots themselves.

Our training centres cannot deliver this at present capacity, and their certificates are not recognised where it counts. Fix the certification, partner with Japanese, Korean and German training bodies, and put English and digital skills inside every course. Beijing has published the timetable. We should stop laughing and start reading it.

The writer is co-founder and CEO of Accfintax