Can AI improve our budget implementation scenario?
On July 25 this year, the Prothom Alo English edition published a news item about a bridge being built somewhere on the outskirts of Dhaka that led to nowhere. Funny as it may sound, such cases are not uncommon. More recently, The Daily Star too ran a story about a bridge lying unused for two decades for lack of any connecting road!
Now imagine the following scenario—a government entity submits a bridge project proposal for approval to the Executive Committee of the National Economic Council (ECNEC), Bangladesh's highest public fund allocation body. The proposal goes through an Artificial Intelligence (AI) powered system. The system points out that the proposed bridge doesn't connect to the existing road network, and the indicated beneficiaries do not match the national demographic database. ECNEC turns down the proposal, saving a significant amount of taxpayers' money from being wasted.
The above scenario may sound like science fiction in the distant future, but it's plausible. Ill-conceived projects are not the only problem though. Several development projects are falling behind schedules or costing more than the allocated budget. Sometimes, a whole ministry is failing to keep up. Such stories come to light only when the Implementation Monitoring and Evaluation Division of the Ministry of Planning (IMED) investigates the progress of projects approved. Other times, a media report brings it to public knowledge. But by then, it may be too late to salvage it. A system based on AI could potentially avoid all these possibilities.
The Bangladesh government has already adopted a national strategy for AI—the National Strategy for Artificial Intelligence in Bangladesh, March 2020. Although it doesn't categorically mention the IMED, there is no reason why this high-level government body shouldn't benefit from such a powerful tool in Annual Development Plan (ADP) implementation monitoring. ADP monitoring is similar to project management; only the scale is much larger. One only has to look at it from an AI perspective to reap the benefits this technology has to offer.
AI can gain insights from past data ("training" or machine learning) and use this knowledge to discover the meaning of new data. As the amount of data increases with time, its knowledge base also grows, making the system more efficient. The quality and amount of the data determine an AI system's effectiveness. On the other hand, biased data will make it error-prone and unreliable. Therefore, the first requirement of an AI implementation system is a vast amount of quality data in electronic formats such as numbers, tables, texts and images. An AI-based system can train itself with these data and produce performance indicators or status reports for future projects after several rounds of refinement.
Government agencies in Bangladesh have been using information and communication management systems for several years. These include e-filing, the Project Management Information System, and e-Government Procurement (e-GP). All these systems gather a large amount of data relevant to project performance, physical progress, budget allocation, fund management and financial progress. Conventionally, it is up to humans to interpret such information, and determine a project's performance and intervention requirement. An adequately trained AI-based system can perform all these tasks and leave decision-making to humans. Of course, the decision-makers won't have to do whatever the system recommends. They can apply judgment as appropriate. Meanwhile, the system will continue working 24/7, analysing large data volumes, and providing time-critical information to support such decision-making.
The other requirement for AI-based systems is a substantial body of technical experts in AI, machine learning (ML) and data science. A close collaboration between such experts and domain expertise is a prerequisite for the success of any AI initiative. Fortunately, Bangladesh has both. Several home-grown software companies are already working on AI-related product development for American or European clients. The IMED will have no problem adding their domain knowledge to such a pool of expertise and start working on a small model. The model can be gradually scaled up as required.
The system can become even more effective with a Geographic Information System (GIS) to store the projects' location information. Bangladesh has been using GIS since the early 1990s on the Flood Action Plan (FAP) projects. Several Bangladeshi software companies are deploying GIS-based analytical tools for overseas customers. The synergy from combining GIS and AI will offer a more powerful tool to the policymakers by pointing out overlaps, gaps and duplication of government investments. Let's examine a few likely scenarios where such a system would be helpful.
A project's physical progress falls behind fund disbursement, and the allocated budget may not be enough for its completion. The system will pick it up and notify IMED to decide on remedial measures. Or, a project director may not be able to mobilise the contractor on time because of the delay in contract signing or land acquisition. It will never be possible for the IMED to identify such an issue with a manual monitoring system unless someone familiar with the project details points it out. But the AI-based system, by continuously interpreting the communications, can highlight it, along with the source of the delay. If any project's location or scope overlaps with that of another, the system will also highlight that. All these will enable the ECNEC members to understand better how the government's interventions are working.
The potential benefits of an AI-based system are aplenty, but there are also many cases of failed AI initiatives. Developing an effective AI-based system needs careful planning, coupled with learning lessons from others' experiences and meticulous work. An IMF blog has warned of a possible AI-induced widening gap between the rich and the poorer nations. But with its efficient and ethical use, AI can close such gaps too. Let Bangladesh launch its fourth industrial revolution with an AI-powered ADP monitoring system and facilitate more equitable distribution of national wealth.
Dr. Sayeed Ahmed is a consulting engineer and the CEO of Bayside Analytix, a technology-focused strategy consulting organisation.