Artificial intelligence may strike Pakistan where it is weakest. A recent World Bank report on global development warns that the technology could deepen the country's job-market crisis. The danger is twofold: automation threatens the low-end outsourcing work behind Pakistan's fastest-growing exports, and it may thin the ranks of Pakistanis employed abroad, whose remittances have long propped up the balance of payments. With some 25m children out of school and public spending on education at around 0.8% of GDP, critics fear the country is entering the AI age with few defences.
The scale of investment elsewhere is staggering. Just five American companies are pouring $775bn into AI, according to one editorial. That spending promises efficiency gains but could wipe out many entry-level jobs, and fresh graduates are already struggling to find work. Fewer educated Pakistanis may secure positions in the West and the Middle East, the regions that supply the bulk of remittances, while the war in the Middle East raises the prospect of an economic slowdown there.
The implications for the external account are serious. For two decades, domestic productivity has slipped relative to the rest of the world, weakening exports and lifting imports in real terms. Strong growth in money sent home by workers has filled most of that gap. If AI curbs hiring abroad while Gulf economies falter, analysts expect Pakistan's external pressures to intensify over the medium term.
Services exports face a similar threat. IT and other business services have roughly doubled to about $6bn over the past few years, but much of that work consists of basic tasks won through cheap labour. AI is rapidly absorbing such jobs at far lower cost than hiring workers in developing countries, eroding that advantage. India, with its vast outsourcing industry, is already bearing the brunt, and Pakistan may soon feel the heat as well.
The disruption is spreading at home too. Small and medium-sized firms are trimming headcounts and subscribing to AI tools that perform tasks once done by staff. Industry experts say many of Pakistan's large IT companies are not ready to move beyond call centres, basic coding and routine number-crunching into more sophisticated work. Moving up the value chain is, as one editorial notes, easier said than done.
Some argue that the deeper problem lies in the classroom. AI, in this view, is arriving not as a magic escalator but as a mirror, exposing rote learning, poor literacy and numeracy, thin research habits and bureaucratic universities. Education, they contend, is the reform that determines whether tax, energy and export reforms can succeed, since each eventually collides with a workforce ill-equipped to reason, adapt or use technology intelligently.
Economists offer a useful lens. Daron Acemoglu and Pascual Restrepo distinguish between technologies that merely replace human labour and those that create new tasks requiring judgment and creativity, with institutions deciding which effect prevails. Acemoglu and Simon Johnson warn of "so-so technologies" that automate without raising productivity. In a memorisation-based system, critics fear, AI will simply produce mediocre work faster and with better formatting.
The starting point is grim. Beyond the out-of-school children estimated by UNICEF, World Bank data on learning poverty show that a large share of late-primary pupils cannot read proficiently. The federal budget for 2026-27 allocates roughly Rs117.7bn in current spending to education affairs and services. Total public spending on education sits far below the 4-6% of GDP benchmark associated with serious investment, a gap one commentator dismisses as fiscal tokenism.
Reformers propose a parallel national learning pathway. It would include community learning hubs, evening and mobile classes, accelerated literacy and numeracy, AI-assisted tutoring in Urdu and local languages, and certification based on competence rather than time spent in class. Older learners would gain skills tied to livelihoods such as agriculture, logistics, healthcare support, freelancing and data work, rather than chasing degrees for their own sake.
Credentials themselves need rethinking; they argue. Generative AI is shortening the shelf-life of many skills, so a degree earned at 22 reveals less about what someone can do at 35. Drawing on work by Acemoglu and David Autor on labour-market polarisation, and Erik Brynjolfsson's warning about the "Turing Trap", advocates call for modular, stackable and renewable certification that lets people return to learning throughout their careers.
Universities come in for particular criticism. Too many, critics say, operate as bureaucratic machines, with leadership chosen through political networks and promotion rewarding seniority and compliance over research and teaching. Building on Nadeem Ul Haque's work on the country's feeble research culture, reformers want genuine academic searches for vice-chancellors, independent boards, protected academic freedom and far more exchange between campuses and disciplines.
Not all the news is gloomy. World Context, a start-up building a data supply chain between Pakistani factories and America's robotics industry, has partnered with Multinet Pakistan to host its growing library of visual data. Its chief executive, Sherdil Ali Khan, aims to produce the country's largest computer-vision datasets. Adnan Zaidi, Multinet's chief, said the deal would link local data capabilities with the global robotics and AI ecosystem, backed by the firm's fibre network.
Officials are also courting new technology. Bilal Bin Saqib, minister of state and chairman of the Pakistan Virtual Assets Regulatory Authority, told the Bitcoin Asia conference in Hong Kong that a regulatory framework for virtual assets was built in under six months for about $200,000, leaving roughly 92% of its budget unspent. A planned Bitcoin strategic reserve, he said, aims to prepare for future finance rather than to amass the currency.
The verdict, then, rests on choices yet to be made. AI could help Pakistan leapfrog stages of education, improve health care and equip underprepared graduates with marketable skills. But the country lacks the buffers to absorb years of economic and employment disruption. Used as decoration atop an unreformed system, the technology will produce better-looking mediocrity; used as a spur to rebuild education, it could prove an opportunity rather than a nightmare.