Will AI work with our demographic dividend?

For many years, Pakistan’s story — particularly, the case for investing — has largely pinned on some variation of the demographic dividend. Basically, the fact that we have a huge youth population of digital natives not only makes us a potential powerhouse as a consumer market but also an attractive source of raw talent. While the argument has merit, it’s not without caveats, which have an inverse relation with optimism.

Traditionally, the counterpoints to the promise of demographic dividend have varied from the substandard quality of the education system to the more macroeconomic explanations, particularly the declining purchasing power. But of late, a new variable, artificial intelligence (AI), has been introduced into the mix, whose direction and quantum of impact are still a topic of debate. It’s not just about Pakistan: advancements in AI have prompted loads of predictions on the potential effect on employment.

On the one hand, you have the techno-libertarians dreaming of a future where all jobs, and even interactions, will be done between agents as the ultimate utopia. Others are more cautious in their aspirations and point to what it would mean for the people, never mind the increase in business productivity. The safer position is obviously that economies can leverage AI to both improve efficiency and keep jobs intact, albeit with a different composition.

Though any and all of these may be true, there is a need to have a better understanding of both the challenges and opportunities posed by AI to Pakistan with regard to both employment and gross domestic product. In economic theory, output is a function of capital and labour inputs, as well as total factor productivity.

Only 12.7pc of our labour force is employed in occupation groups that can be deemed to have exposure to AI

Broadly speaking, there are potentially two opposing forces at play. While technological advancements such as generative AI, typically represented by total factor productivity, should help boost output, they will also reduce the need for labour inputs through automation. For now, their magnitude isn’t exactly clear, but existing literature gives us some indication of which areas are prone to AI adoption.

By and large, there is a consensus among economists and researchers that the most recent wave of AI will impact the service sectors the most — a fact that all of us ‘knowledge workers’ have firsthand experienced, whether for research, data analysis or any other task. As per the Organisation of Economic Cooperation and Development, telecom, media, and IT services stand out in AI across four dimensions, ie human capital, innovation, exposure and usage.

In this regard, Pakistan’s sector-wise composition may give a sense of relief, considering that more than a third of our labour force is engaged in agriculture, forestry & fishing, as per the 2020-21 survey. Construction, which employs 9.5pc of the people, is also believed to be less exposed to genAI.

Meanwhile, services are likely to be most vulnerable and employ 37.5pc of the labour force, though the share of ‘knowledge workers’ is understandably much smaller. Another approach to gauge the impact is to look at the employment composition of occupation groups. Generally, advanced economies tend to have a higher share of workers in professional, managerial, and technical occupations with high AI exposure, which can either help augment their capabilities (such as professionals or managers) or replace them (eg clerical support).

This is where Pakistan’s advantage of backwardness manifests itself: only 12.7pc of our labour force is employed in occupation groups that can be deemed to have exposure to AI. Of that, professionals and managers make up more than half (7.5pc) and have a complementarity factor. In other words, they are likely to be augmented, not displaced, by AI.

Unfortunately, there is a lack of literature on the exact magnitude in the local context. However, we can take India as a peer as the underlying employment structure is somewhat similar. There, the impact of AI is supposed to be quite minimal due to a smaller share of people occupied in ‘knowledge’ work.

However, the sense of relief is at least partially misplaced, given Pakistan’s long-term productivity challenges. At around 1pc, the country’s mean growth in GDP per worker between 1992 and 2001 has been lower than all our peers and below the South Asian and lower-middle income averages. So even if the impact of AI may not be as high in terms of jobs automation, there’s no denying the fact that our labour force is already way behind the curve in even quasi-natural intelligence.

Similarly, the consequences can be dire within small pockets where IT and other business services are a great case in point. Both of them, particularly the latter, are important sources of forex inflows and almost entirely rely on labour arbitrage to deliver low-value work and therefore, are highly exposed to AI. While the Labour Force Survey gives no such breakdown, these two sectors employ around a million people and are already fighting a race to the bottom approach in terms of costs against other cheap markets. How will they compete against machines?

The writer is the co-founder of Data Darbar and works for Karachi School of Business and Leadership

Published in Dawn, The Business and Finance Weekly, May 5th, 2025