Bangladesh Needs More Than AI Skills

It needs institutions capable of turning technology and human talent into productivity.

AI-generated editorial illustration.

Two recent articles in The Daily Star appear to address different problems.

The first reported on a roundtable calling for Bangladesh to move beyond digital access and build national capability in artificial intelligence. Speakers emphasized education reform, AI skills, university-industry collaboration, and an ecosystem for local innovation.

The second examined Bangladesh’s brain drain. It described researchers financing their own work, graduates unable to use their training, and grassroots inventors receiving institutional attention only after becoming visible on social media.

These are not separate problems. They are two sides of the same institutional weakness.

Bangladesh does need people who understand AI and other emerging technologies. But producing more skilled people will achieve little unless businesses, universities, and government agencies can employ their capabilities productively.

The larger challenge is not simply developing talent or acquiring technology. It is building institutions capable of turning both into results.

The debate is too narrow

The case for investing in AI skills is not unreasonable.

A more capable workforce can pressure companies and public institutions to modernize. New skills can create new businesses, improve existing ones, and allow Bangladeshis to compete in global markets. Without investments in education, technical competence and digital infrastructure, Bangladesh risks falling further behind.

But skills supply and institutional demand must develop together.

Teaching thousands of young people data analysis, programming, or machine learning does not automatically create productive employment. Someone must have a real problem for them to solve, the authority to employ them properly, and the management capability to turn their work into measurable value.

Otherwise, Bangladesh may produce more qualified people without producing enough serious places for them to work.

Technology is more than AI

Public discussion increasingly treats “technology” and “AI” as though they mean the same thing. They do not.

For many Bangladeshi businesses, the next major productivity improvement will not come from developing an AI model. It may come from better machinery, preventive maintenance, production planning, quality control, inventory management, industrial automation, or a properly implemented accounting or customer-management system.

A factory that does not know its machine downtime, rejection rate, or actual production cost is not ready for an AI transformation. A company that installs an ERP system but continues managing the business through disconnected spreadsheets has not meaningfully adopted technology.

Technology creates value only when it changes how an organization operates.

In my own work in industrial services, access to technology is rarely the only constraint. Modern machinery, software, and control systems can usually be sourced. The harder work is selecting the right solution, installing it properly, maintaining it, training people to use it, and holding someone accountable for the outcome.

That is as much a management challenge as a technical one.

The World Bank Group’s Bangladesh Country Private Sector Diagnostic, published in April 2025, makes a similar broader point. It identifies energy shortages, weak corporate governance, financing constraints, and an uncertain business environment as barriers to private investment and productivity. It also notes a substantial mismatch between graduate capabilities and employer demand.

AI training cannot solve these problems by itself.

Skills without institutions become brain drain

Bangladesh often responds to unemployment and low productivity by launching another training programme, certificate course, or skills initiative.

Training can improve individual capability. It does not guarantee that organizations will know what to do with that capability.

A technically strong graduate may join a company that recruits through connections, resists new ideas, or assigns skilled employees to routine administrative work. A researcher may have knowledge but no dependable funding, equipment, or institutional support. An inventor may build something useful but have no credible path to engineering assistance, intellectual-property protection, or commercialization.

Management weakness makes the problem worse.

Someone must define the problem, redesign the process, choose the technology, assign responsibility, enforce adoption, and measure whether performance improves. Without that layer, companies purchase systems they do not use, machinery they do not maintain, and training they do not apply.

The result is not merely wasted investment. It is wasted talent.

When capable people repeatedly encounter institutions that cannot recognize, use, or reward their abilities, underemployment becomes frustration. Over time, leaving the country begins to look less like ambition and more like the only rational career decision.

This is why brain drain cannot be addressed only through patriotic appeals or attempts to persuade people to remain in Bangladesh. The country must give capable people serious institutions in which to do serious work.

AI should solve problems, not become the programme

Bangladesh should invest in AI, but selectively.

There are areas where local capability could create real value: Bangla language tools, speech recognition, optical character recognition, agricultural advice, flood management, healthcare administration, industrial analysis, and better delivery of public services.

Bangladesh should also be able to evaluate imported AI systems for accuracy, bias, safety, and performance in local conditions.

But the country does not need to imitate wealthier nations by pursuing frontier models or expensive infrastructure simply because AI has become strategically fashionable. Most near-term value will come from adapting existing technologies to specific Bangladeshi problems.

The proper sequence is straightforward:

Identify an important problem. Improve the underlying process and data. Select the appropriate technology. Train the people responsible for using it. Measure whether the result improved.

Starting with “we must use AI” reverses that sequence.

It creates a risk of conferences, training targets, pilot projects, and technology centres whose success is measured by how much activity they generate rather than whether anything works better.

AI cannot manufacture accountability

The roundtable also suggested that AI could improve government monitoring, transparency and accountability. It may indeed help identify irregularities, analyse projects, and reduce processing delays.

But AI cannot create accountability where the institutional foundations are absent.

A system cannot reliably analyse incomplete or manipulated records. An automated warning has little value if officials can ignore it without consequence. A digital decision is not accountable unless affected citizens can understand it, challenge it, and have errors corrected.

Responsible government use of AI therefore requires reliable source data, clear ownership, audit trails, privacy protection, human review, and effective appeal mechanisms.

Otherwise, Bangladesh may automate poor processes and make responsibility even harder to locate.

Governance must come before government AI. It cannot be expected to emerge from the technology itself.

Start with adoption, not ambition

A practical national strategy should begin with the foundations that allow technology to work: reliable electricity, affordable connectivity, access to finance, predictable import rules, cybersecurity and usable public data.

The next priority should be helping existing firms adopt proven technology. This includes machinery, production systems, quality control, energy efficiency, maintenance, financial systems, and digital management tools. Public support, where justified, should be tied to implementation and measurable productivity improvements, not simply the purchase of equipment or completion of training.

Education policy must then be connected to this demand. Companies should help design courses, offer apprenticeships, and define the practical capabilities they require. Universities should be judged partly by whether their graduates and research solve real problems, not only by enrolment, certificates, and publications.

AI should sit within this broader system as a tool for selected problems, not as the organizing principle for the entire technology agenda.

Begin with one sector

Bangladesh does not need another strategy that attempts to transform everything at once.

It could begin with one sector, such as light engineering, and select a small group of firms willing to participate seriously.

Measure their current productivity, machine downtime, defects, energy use, delivery times, and employee capabilities. Help them introduce appropriate machinery, maintenance systems, quality controls, and digital management tools. Train employees around the technology actually being installed. Measure the results after twelve months and publish what worked, what failed, and why.

Successful methods could then be adapted for garments, agro-processing, pharmaceuticals, logistics, and other sectors.

This would be less dramatic than announcing a national AI ecosystem. It would also teach Bangladesh something far more useful: how to convert technology, talent, and management into measurable productivity.

Bangladesh’s challenge is not a shortage of technology, ideas or talented people. It is a shortage of institutions that know how to use them.

Until that changes, the country will continue training people for opportunities that exist mainly somewhere else.