Interviewed by Syed Md. Rakeen, Team MBR

Mohammad Oli Ahad is the founder and CEO of Intelligent Machines, which builds AI products to help clients reach their organisational goals and objectives. After graduating from the Institute of Business Administration, University of Dhaka, he started working for British American Tobacco, where he led global and regional projects for 25 countries. He also worked for Grameenphone, Holcim, and the International Fund for Agricultural Development (a specialised UN agency). Team MBR was in a conversation with Mr. Mohammad Oli Ahad and was fortunate enough to receive his take on the prospects of AI in Bangladesh.

Syed Md. Rakeen: Some of the businesses in Bangladesh have already been able to automate certain operations, save costs, and enhance efficiency owing to the successful implementation of Artificial Intelligence (AI) technology. Would you kindly share your views with us regarding the extent to which Bangladeshi firms have explored the potential of AI?

Mohammad Oli Ahad: In recent years, a handful of forward-thinking Bangladeshi companies have started successfully implementing AI, with great results. bKash has gained 288% more productivity in serving 680,000 outlets with only 1,400 merchandisers. Unilever has been able to reach more than 10% of their target results for their brand campaigns in scope using fewer resources than they used to spend to reach only 1% previously. IDLC Finance has been using AI to verify and analyse bank statements as well as the central bank’s CIB reports and calculate customers’ credit scores, bringing down the processing time from days to minutes. A globally leading consumer goods manufacturer is piloting AI to optimise their spare parts’ replacements and also working to prevent counterfeit products with AI.

What is greatly encouraging is that these and other Bangladeshi firms’ adoptions of AI are in diverse industries and serve a wide spectrum of important functional purposes. However, decision makers at most organisations still tend to shy away from undertaking AI projects, considering this a highly specialised category and feeling inadequately prepared.

This is more of a behavioural factor than anything to do with the technology. The key to successful projects is the ability to ask the right questions and a deep understanding of the business. Winning organisations already possess these key ingredients to solve their high-value problems with AI, but they will need to overcome their hesitations. A way to do that could be to know more about the processes and risks of AI projects.

Syed Md. Rakeen: The Fourth Industrial Revolution (4IR) will definitely change the way most of Bangladesh’s industries work. What key initiatives do you suggest to make Bangladesh prepared to exploit the bright sides of 4IR?

Mohammad Oli Ahad: 4IR adoptions are still disproportionately concentrated on a small section of the world, and a new form of inequality is developing right now, which is greatly alarming. We recommend three immediate sets of actions. Firstly, encouraging and facilitating Bangladeshi firms to identify scopes and start adopting 4IR technologies without further delays. Secondly, building the desired skills immediately on a large scale. For instance, we believe that Bangladesh can produce 2,000 highly capable AI engineers in the next two years without any major trouble and can become a defining force globally in this regard. Finally, establishing specific goals and ensuring that progress is measured and tracked again, without any delay.

The truth is we need to do this not to ‘exploit the bright sides’, but to ensure we are not on the way to becoming the new age’s devastated batch of economic colonies by rapidly progressing developed markets. This race is quickly becoming like what we have come to see in the Olympics. Only a handful of countries will continue to dominate the leaderboard, while the rest of the countries will only play futile catch-up. We must act immediately.

Syed Md. Rakeen: A report published by the International Monetary Fund (IMF) in 2022 stated that around 60% of the labour force in Bangladesh is employed in industries that are at high risk of automation. What are the key skills individuals need to acquire to prevent becoming obsolete in the job market?

Mohammad Oli Ahad: The first set of skills we will highly recommend are behavioural. They are building self-awareness, developing a constant learning habit, and growing a resolute drive for results.

The second set is technical skill, which also comprises three elements. They are,

(i) Solid grasps of key concepts in any of the core STEM branches, such as statistics, physics, et al. A reassuring fact is that while the ‘top layers of application knowledge’ like ChatGPT, TensorFlow, or PyTorch would likely be replaced every few years (the next batch of architectures will always be in the cooking!), core functional knowledge is likely to remain useful at least for the next two to three decades.

(ii) Time-relevant practical technology skills such as continuous improvement and continuous development (CI/CD), event streaming like PubSub and as such, not just basic coding. What is it that companies need now to deploy advanced technologies consistently and at scale? Those skills.

(iii) And differentiating skills in the liberal arts such as experience engineering, design thinking, et al.

It is important to realise that we cannot now plan to catch up by learning past decades’ skills that we have missed (such as basic coding) to aim to outsmart rapidly progressing developed economies that are constantly working at the frontier technologies of tomorrow.

And the third set would be developing deep domain expertise. No amount of technology skillsets would be of much use if we did not understand the business and market really well. Deep expertise in finance, consumer behaviour, et al., would be just as crucial.

While developing these skillsets, we will also need to plan to support those who would be adversely affected on a large scale and may not be in a position to support themselves and their dependents for near to mid-term time periods. Otherwise, we run a great risk of grave-scale human catastrophes. Large demographics will face questions of sustenance and might fall far behind.

Syed Md. Rakeen: According to the CISCO Digital Readiness Index 2021, Bangladesh was ranked 109th out of 146 countries, behind its South Asian neighbours India, Nepal, and Sri Lanka. What measures can the government take to reduce the gap with its neighbours?

Mohammad Oli Ahad: The two areas that could get us maximum results in a reasonable timeframe are also very challenging, as they entail great political willpower.

(i) Building and Fixing the Technology Infrastructure: We keep seeing reports that our telecom and internet bandwidth are the most expensive both globally and regionally. We must choose to fix these and other infrastructure issues if we want to move up in this area.

For instance, we do not have supercomputer infrastructure for tech teams. As of now, teams in Bangladesh need to rent supercomputers from abroad at exorbitant rates, with a markup of more than 300%. Without supercomputing capabilities, how would we develop and run AI solutions? In our country, we have been focusing only on primary cloud storage to date, not on the computational capabilities. This will forever remain a far-fetched game without the required capabilities.

(ii) Establishing Accountability at All Levels of Government and Private Organisations, Including Law Enforcement and the Judiciary: No matter what we do, if we cannot bring in the required accountability (i.e., where one can expect the due services without any undue expectations by the operating agencies), it will remain one of those arithmetic problems where you have one faucet to fill a bucket with three holes in it.

So, at the end, it boils down to long-term willpower to bring in the desired changes, which is a very, very uphill climb, no doubt.

Syed Md. Rakeen: Experts suggest that investments in AI can generate up to a tenfold return. Would you kindly share with us the challenges Bangladeshi firms may face while implementing AI? What strategies can they follow to maximise their return on investment (ROI) with the help of AI?

Mohammad Oli Ahad: In our calculation, we see not a tenfold return but a staggering 32- fold return. Bangladeshi companies that have successfully implemented AI for their high-value problems gained at least 32 dollars within the first six months for every dollar spent on AI projects.

The challenge? The irrational fear of decision-makers to undertake AI projects. Most are going through a hesitant period, unsure of their own capabilities to manage AI projects successfully. As a result, they continue to perceive the risks of failure as being higher than the missed opportunities, which cannot be farther from the truth!

The truth is that capable managers possess the key ingredients to apply AI successfully to their high-value problems, and it is their deep understanding of the business and their capability to guide the AI engineers by consistently asking the right questions that make this possible.

A winning strategy could be to divide AI projects into four phases: feasibility (3 weeks, generally complimentary), POC (2-4 months at a cost of USD 8,000-18,000 where they will get the AI model), an optional phase of pilot (2-3 months at a cost of USD 3,000-4,000 where they will get live results for the AI model), and planning for the go-live in 1 month with a budget of USD 3,000 per month. This is for a high-value, single-task, complex AI model.

We believe organisations could begin by forming a spearhead team with the right mindset. This team will meet with tech teams and talk with other enterprises that have already applied AI in Bangladesh. Armed with this knowledge, this team would then select one high-value problem for the organisation and present a plan to the management team within two weeks of its formation.

The organisations will need to ensure they have all four key elements of a successful technology project: a budget of USD 10,000-25,000, timeline of one to three quarters, and authority to approve the project.

Syed Md. Rakeen: ChatGPT has taken the world by storm, gaining 1 million users in less than a week since its launch. What impact can Natural Language Processing (NLP) models like ChatGPT have on businesses in Bangladesh?

Mohammad Oli Ahad: Plenty if the business models could fit right. But first, we will have to be aware of two things. Firstly, ChatGPT’s sole service is finding the next right word from its base data, and that is it. So, we will likely need additional work to ensure it provides consistent, safe, and compliant responses. And, secondly, ChatGPT or any other large language model would also incur computation costs. So, it would only be feasible if the cost justified high-value use cases where such digital agents could provide good value.

Early targets could be agent-based services such as bank tellers; bus, train, and other ticketing services; healthcare operators to book appointments and tests and share the results; call centres; and personal assistants in a wide scope. Also, execution jobs that could be wholesomely based on previous records, such as those of content creators, legal and secretarial assistants, social media listening tools for managing brand perceptions, et al.

Syed Md. Rakeen: To ensure a smooth transition to an AI-driven economy, the educational curriculum at every level in Bangladesh is required to be redesigned. What are some of the steps you think can help students equip themselves with the necessary skills for tomorrow?

Mohammad Oli Ahad: We already have a good understanding of this roadmap.

Firstly, we need to begin by enlightening students on the different roles in AI. To illustrate, consider healthcare or many engineering fields. There are many professional roles in these fields: surgeons, nurses, diagnosticians, pathologists, et al., and we generally know which path to follow to gain the key skills in these fields. As of now, students largely do not have such clear ideas about AI or data science. Such as, what are the different roles in AI? What strengths do they bring in, and what purposes do they serve?

Next, we need to share what would be required to perform at a high level in those specific roles. What would be required for a level-2 AI engineer’s role, and in level-3? Or, what would be required for a level-2 data scientist’s role?

And, then, we will need to provide them with a large real-life dataset to practise their skills. Similar to how doctors practise during their internships at medical colleges dealing with real-life patients, we will need to provide them with an opportunity to practise their skills. Otherwise, learning only with online resources may not help develop the deep expertise that will be needed.

Finally, we will need to provide them with an arrangement to continuously assess and track their progress.

Syed Md. Rakeen: The mass implementation of AI carries the risk of it being used for unethical purposes. What ethical aspects should be taken into consideration while implementing AI, and what measures should be in place to prevent unethical practises?

Mohammad Oli Ahad: This is a high-impact risk with a high likelihood. It is very real. We will need to start holding technology companies accountable for their actions’ impacts. They cannot be allowed to make large sums of money without taking any responsibility for the way their tools are being used.

So, we will need our legislators to be aware of and educated about these impacts in the hope they will enact the legislation. We will also need third-party auditors to run routine assessments and share public reports. We must also discuss this with our children, both at home and at school.