A RETAIL brand in Dhaka may use ChatGPT to write a Facebook post. A restaurant group in Sylhet may use AI to reply to customer reviews. An exporter may use it to draft emails for foreign buyers. These are useful examples, but they do not mean a business has fully adopted AI.
AI use in Bangladesh is still at an early stage. For many people and businesses, it is limited to casual uses such as ChatGPT, image creation and basic content writing. But the draft National AI Policy 2026–2030, released in February 2026, points to a bigger goal. It wants Bangladesh to move beyond casual AI use and apply AI to improve productivity, support better decisions, use data more intelligently, encourage local innovation, protect people from AI-related risks and make the economy more competitive.
The Bangladesh AI Readiness Assessment Report, published by UNDP in November 2025, welcomed the country’s move towards an AI policy framework. But it also made one thing clear: readiness is still uneven. Bangladesh needs stronger connectivity, better skills, sound data governance and responsible use of AI. This matters especially for small and medium businesses. Before they adopt AI at a higher level, they need basic readiness: a clear roadmap, trained staff, privacy rules and human review of AI outputs. AI should support decisions, not replace human judgement.
For businesses, AI adoption can be understood in five simple stages. The first is awareness. At this stage, a business has heard about AI but does not clearly understand what it can do or how it can help.
The second stage is casual use. A business may try tools like ChatGPT for general questions or simple ideas, but there is little or no direct link with business operations.
The third stage is productivity use. AI starts helping with regular tasks such as writing emails, preparing product descriptions, creating social media posts, translating messages or drafting customer replies.
The fourth stage is decision support. Here, AI is used to collect, organise, compare and analyse information. For example, a business may use AI to review customer feedback, identify common complaints, compare competitors, summarise sales trends or find areas for improvement. At this stage, AI supports decisions, but humans still make the final call.
The fifth stage is intelligent integration. AI becomes part of business systems and workflows. It may support marketing campaigns, customer review responses, inventory alerts, customer segmentation or routine task automation. AI is no longer used only when someone opens a tool. It becomes part of how the business operates.
The danger for many businesses is to confuse activity with progress. Writing a few social media posts with AI may save time, but it will not change the business. Real value comes when AI helps a business understand customers, improve service, reduce waste, compare options or make faster, better-informed decisions. This is where the gap will grow between businesses that only try AI and businesses that use it with discipline.
From my own work, I have seen how AI can support both decision-making and controlled automation. I use AI to gather information, compare offers and identify the most suitable option before making a decision. I also use AI to automate limited tasks under clear rules. For example, positive customer reviews can be answered automatically, while negative reviews are drafted by AI but reviewed by a human before they are published. In restaurant operations, I have found this especially useful because reviews arrive daily, and not every review needs the same level of attention. This is the right balance: automate low-risk tasks, but keep human judgement for sensitive decisions.
Moving from one stage to another does not require expensive technology immediately. It should begin with one real business problem. After choosing the problem, the business should train its people to use AI properly, check AI outputs and protect sensitive information. It should also set simple rules on privacy, accuracy and human approval.
AI adoption must also be responsible. A business should not treat AI outputs as automatically correct. AI can produce wrong information, weak analysis or confident but misleading answers. It can also create privacy risks if staff upload customer data, financial details or internal business information into public tools without control. Sensitive data must be protected, important outputs must be reviewed and final decisions must remain with humans. The best use of AI is to combine machine speed with human judgement.
The question for Bangladeshi businesses is no longer whether AI exists. The real question is: where are they now and what must they do to move one stage higher responsibly? If AI does not improve real work, it remains a novelty. AI adoption is not about using AI tools. It is about building business capability.
ABM Kamrul Huda Azad is director of Yaki Ya, a Japanese restaurant chain operating in the UK and Middle East, and head of operations, UK and Canada, at Daytona Capital Management, UK.
