Walk into a modern factory, hospital or bank, and nobody boasts about using electricity. A century ago, it was a technological marvel exhibited at world fairs. Over time, it became indispensable, transforming agriculture, manufacturing, healthcare and communications.
Artificial intelligence is following a similar trajectory.
When computer scientist and Coursera co-founder Andrew Ng described AI as “the new electricity”, he was identifying a technology with the potential to transform almost every industry.
Over the next decade, AI will become less of a standalone application and more of an underlying layer across digital systems. Conventional software follows rules written by humans: click a button, execute a command. AI allows software to identify patterns, predict outcomes and adapt to changing conditions.
Smartphone cameras already use neural networks to adjust lighting before a photograph is taken. Similar technology will increasingly manage urban traffic, detect crop diseases from farmers’ photographs, identify fraudulent financial transactions and support medical decisions using patients’ genetic information.
AI could become as integral to software and industrial machinery as electricity is to modern appliances. But access to electricity means little without the skills to use it. The same applies to AI.
AI literacy is becoming a basic professional skill, alongside reading, writing and digital competence. Yet many people still treat generative AI as a novelty or a substitute for thinking, using it to draft routine emails or summarise documents they have not read.
That barely scratches its potential.
Productive AI literacy means understanding how to use these systems to solve problems, improve decisions and increase efficiency.
Software engineers can use AI coding assistants to develop and test applications faster. Teachers can prepare exercises tailored to individual students. Small businesses can forecast inventory demand, analyse customer feedback and manage accounts without expanding administrative teams.
The gains depend on knowing what to ask, how to assess the answers and where to integrate AI into existing work. Workers who develop these skills will have an advantage over those who do not.
For Bangladesh, however, individual competence is only part of the challenge. The country also needs AI infrastructure capable of serving its language, institutions and economy.
Most leading large language models (LLMs) are trained predominantly on English-language material. English can account for as much as 95% of some pre-training datasets, leaving thousands of other languages competing for a small share.
Bangla, despite being spoken by hundreds of millions of people, remains underrepresented in many mainstream training datasets, sometimes accounting for less than 0.5%. The consequences extend beyond linguistic representation.
AI models rely on tokenisers, which divide text into smaller units for processing. Models that are poorly optimised for Bangla script may require more tokens to process the same information, increasing computational costs and reducing efficiency.
There are also problems of accuracy and context.
A foreign model processing a Bangladeshi land deed may struggle with administrative terms such as Mouza, Khatian and Daag Number. It may generate plausible but incorrect explanations of legal procedures.
Models trained primarily on Indian Bengali can introduce differences in vocabulary, usage and cultural references, substituting Jol for Pani, for example, or using forms of address less common in Bangladesh.
Such differences matter when AI is deployed in public administration, legal services, education and healthcare. Bangladesh therefore needs to develop its own Bangla-language AI capabilities.
This is not simply a question of cultural identity. It concerns technological sovereignty, public-service delivery and economic competitiveness. If Bangladesh relies entirely on foreign models to process its laws, administer public services and develop digital products, it will remain dependent on systems designed and controlled elsewhere.
The country does not need to reproduce every advance in AI research. Open-source models and publicly available research already provide a foundation for domestic development.
Nor is engineering talent the principal constraint.
Bangladesh however lacks sufficient investment in computing infrastructure, high-quality Bangla datasets and coordinated research capacity. A national programme should aim to develop a Bangla foundation model capable of understanding the country’s legal system, administrative procedures, history, literature and linguistic diversity, including Sylheti, Chittagonian and Rangpuri.
That would provide a common technological base for universities, startups and public institutions to develop specialised applications.
The potential uses are substantial.
An AI-powered tutor could provide personalised educational support to millions of students. Voice-based medical assistants could help doctors in understaffed rural clinics. Government agencies could automate routine administrative work, improve document processing and make public information more accessible.
These applications would require rigorous testing, safeguards and human oversight, particularly in healthcare and public administration. But the potential productivity gains justify serious investment.
The economic case is equally compelling.
Global competition is increasingly shaped by the ability to develop and deploy AI. Countries that build this capacity can improve productivity, reduce operating costs and create higher-value industries.
Those that fail to do so risk becoming dependent consumers of foreign technology. Bangladesh’s existing growth engines are already exposed. The ready-made garment industry, which dominates the country’s exports, competes with manufacturers adopting AI for quality inspection, demand forecasting and supply-chain management.
Factories that use these systems effectively can reduce waste, improve production planning and respond faster to changing consumer demand.
Bangladesh cannot assume that low wages alone will preserve its competitive position. The same pressure is emerging in digital services.
Hundreds of thousands of Bangladeshi freelancers earn foreign currency through software development, graphic design, content writing and customer support. AI is changing the value of this work.
Tasks that once required hours can increasingly be completed in minutes. Clients have less reason to pay a premium for routine digital services when automated alternatives are available. Demand is shifting towards professionals who can combine technical expertise, judgment and AI tools to deliver more sophisticated work.
Now, for Bangladesh’s freelancers, the choice is not between using AI and avoiding it. It is between moving into higher-value services and competing in markets where basic tasks are becoming cheaper. The same principle applies to the wider economy.
Delays in developing domestic computing capacity, Bangla-language datasets and practical AI education will widen the productivity gap between Bangladesh and faster-moving competitors.
So, the country needs a coordinated strategy involving government, universities and the private sector.
Public investment should support shared computing infrastructure and high-quality language datasets. Universities should strengthen AI research and technical training. Businesses should invest in practical adoption rather than treating AI as a marketing exercise.
Education policy must also change. Students and workers need to understand not only how to operate AI tools but how to question their outputs, recognise errors and apply them responsibly.
These investments will not deliver immediate returns. Building reliable language models, training skilled workers and integrating AI into institutions will take time.
But postponement carries its own costs. Countries that master AI will be better placed to modernise public services, improve healthcare and education, and export higher-value technology. Those that remain dependent on imported systems will face recurring costs, limited control over critical infrastructure and weaker prospects for domestic innovation.
Bangladesh has spent decades expanding access to electricity and digital connectivity. The next challenge is to build the intelligence infrastructure that will run on top of them.
The country must decide whether it wants merely to consume AI or develop the capacity to shape and control its use.
—
Syed Almas Kabir is the Chairman of Bangladesh ICT & Innovation Network (BIIN)