On October 27, 2026, the Minister for Parliamentary Affairs told the Senate that Pakistan’s national AI ecosystem was being built steadily, citing a billion-dollar investment target by 2030 and Pakistan’s accession to the newly formed, Shanghai-headquartered World AI Cooperation Organisation (WAICO) as a founding member. Eight months earlier, implementation of Pakistan’s National AI Policy had stalled: the provinces had never formally responded to requests for input, the composition of the AI Council itself was under dispute, and of the policy’s four pillars, only “Awareness and Readiness” showed any real activity. Both statements can be true at once: it is the gap between them, not either claim by itself, that warrants measurement and scrutiny.
Three ordinary moments capture this conundrum better than any white paper could. You could be filling out a government form online, only to be asked for documents that a provincial office already holds. You could receive a NADRA biometric verification SMS without any way of knowing what happens to the data behind it, or whether anyone even checks. You could be booking a Careem ride or placing a Foodpanda order and trusting, on faith, that the person on the other end is being paid and treated fairly, because no one publishes the figures either way. Each moment sits on top of a governance question Pakistan is actively trying to answer: how artificial intelligence should enter public administration; how digital government ought to be judged a success; and who, if anyone, is meant to regulate the platforms that already run a large part of daily life and have become somewhat indispensable for many Pakistanis.
Pakistan is racing to build AI policy, digital-government platforms, and gig-economy rules all at once, without building any way to test whether any of it is working – while China, however imperfectly, already has begun building the relevant instruments. As Islamabad legislates, digitises and narrates its way through three governance frontiers simultaneously, it has yet to ask the question Beijing’s own researchers have started asking of themselves: is any of this actually succeeding?
Building Without a Feedback Loop
Anyone who has restarted an online application because two government apps, systems or platforms failed to recognise each other already understands the problem long before anyone writes about it: Pakistan’s digital government rollout – provincial portals, federal databases, biometric re-verification drives – has multiplied the channels through which citizens meet the state, but without multiplying the ways the state finds out whether citizens are satisfied with what they meet.
In Chinese scholarship, this exact question has a name – “user evaluation” and “public satisfaction” research – and it has moved through two generations in barely five years. The first generation simply asked whether citizens were satisfied with a given portal or app, usually through surveys feeding structural equation models. The second, more recent generation treats satisfaction less as a finish line and more as a gauge: a 2025 study using the XGBoost machine-learning algorithm models satisfaction with local digital public service policy directly from behavioural data rather than self-report; a companion three-dimensional institution-subject-public-perception framework tries to evaluate digital informatisation on more than one axis at once; and a separate strand connects digital government construction to civic policy participation and, more ambitiously, even to measurable gains in environmental attention and pollution reduction.
Satisfaction, in simpler words, is treated as a mechanism that produces outcomes rather than an opinion poll that ends the inquiry. Genia Kostka’s recent work on the expanding digital reach of the Chinese state, and her study with Haili Li on citizens’ engagement with local mobile-government platforms, tie satisfaction to trust more carefully still. A 2025 study by Xing Chen and Hanchen Jiang finds that digital government builds citizens’ trust in local government through institutional design and procedural transparency rather than through service outcomes alone – a mechanism that sits alongside, rather than replaces, the satisfaction metrics above. A provocative study published in 2026 goes further, asking whether digital surveillance itself can manufacture a citizen compliance that merely resembles satisfaction – a distinction worth keeping in mind before any Pakistani official celebrates any uptake figure or tech-adoption statistic.
As Islamabad legislates, digitises and narrates its way through three governance frontiers simultaneously, it has yet to ask whether any of this is actually succeeding or not
Pakistan has almost none of this measurement architecture yet, though it has only just begun. A 2025 systematic review of twenty studies on digital trust and e-government adoption in Pakistan found that research on strategies that build trust, as distinct from simply measuring trust as a variable, remains thin on the ground: the government’s own Citizen Satisfaction Index, run through the Open Government Partnership, is a genuine step but still a narrow one. It represents a start, not an architecture – in its current form, it resembles a survey more than a system.
More promising, and considerably more recent, is the National Data Governance Policy unveiled in June 2026, which creates a Pakistan Digital Authority, a National Data Governance Council, a Chief Data Officer in every public body, and, on paper, a “National Data Maturity Index”. As of the submission of this report, it is still pending approval from the federal cabinet and subsequent notification in the Gazette of Pakistan. Whether it becomes a working measurement instrument or another index that exists mainly inside a policy document, that is exactly the kind of question Pakistan has no settled way of answering – because the apparatus to answer it does not exist either.
Who Audits the Algorithm?
Booking a ride or a delivery order has become so routine that almost no one asks the obvious regulatory question: who, if anyone, is checking whether the person on the bike or in the car is being paid fairly, working reasonable hours, or protected when an algorithm that assigns jobs decides, without appeal, that they are no longer wanted? Careem, Foodpanda, inDrive and Bykea have become part of how Pakistan’s cities function; what has not kept pace is any institution whose job is to monitor and evaluate them against the words “fair”, “safe” or “transparent” in anything more than name. Anyone who feels manipulated – or has actually been cheated – by any of these services ought to ask this question while lamenting the obvious absence of consumer rights protection in Pakistan; the latter is one thing that both the government as well as private sector entities can learn and more readily adapt from China, if they are even remotely serious in implementing it and introducing a modicum of improvement in Pakistani citizens’ daily lives. But this is an end-user quality assurance argument.
Pakistan’s digital-government rollout has multiplied the channels through which citizens meet the state, but without multiplying the ways the state finds out whether citizens are satisfied with what they meet
Coming back to the service-provider-as-employer side of the equation: in regulatory terms, a framework and a monitoring-and-evaluation mechanism are missing here. It is essentially a system of indicators, dynamic data-sharing requirements and periodic review, rather than a rulebook written once and left to enforce itself – which is unfortunately the case with most of the laws and regulations enacted in Pakistan. China offers the clearest illustration of how such a mechanism actually gets built, especially because it shows the stages a regulator has to pass through.
Between 2020 and 2022, Beijing ran what one assessment by Yiping Huang called “strong regulations” – a campaign-style crackdown on the platform giants, examined in detail by researchers tracking the ‘domain-specific centralisation’ that followed. That phase has since given way to calibrated tooling: a 54-article draft proposal released by seven government agencies in October 2026 requires platforms to explain their algorithms, prohibits the use of “abnormal data” as a performance benchmark, and forbids automated decisions that significantly harm a worker’s interests, on top of extending minimum-wage and social-insurance coverage. In an earlier move, a 29-article regulation issued jointly by three regulatory bodies in December 2025 banned e-commerce platforms from forcing lowest prices or abusing algorithms outright. Underneath both the 2025 regulation and the 2026 proposal sits an increasingly quantitative monitoring layer: the “PMC index” used to evaluate digital economy policy, as well as growing academic literature on the uneven enforcement of China’s platform rules across provinces and on what one 2026 study by Qian Li calls China’s “top-heavy” model of digital platform governance. All of this evidence shows that even a centralised state finds real-time measurement difficult, not easy.
Pakistan is several stages behind, and has mostly been diagnosing the gap rather than closing it. The Senate passed a law in January 2026 creating a formal registration, licensing and data-sharing framework for ride-hailing – but only for the Islamabad Capital Territory, leaving unregulated the provinces where most ride-hailing and delivery work actually happens. Elsewhere, the record reads like a catalogue of absence: LEAP Pakistan calls it a legislative vacuum outright, and separate reporting has tracked how the sector’s growth has consistently outpaced its legal protections; the second Fairwork Pakistan report found that none of the country’s platforms meet basic fair-work thresholds; and the first-ever official count of gig work, drawn from the Pakistan Bureau of Statistics’ Labour Force Survey, only arrived in 2026. This means that Pakistan has only just begun measuring the size of the sector it is being asked to regulate, let alone how fairly it treats workers and consumers: the people inside it, and the people these sectors are being built to serve, respectively.
AI as Subject and Narrator
Every government that adopts artificial intelligence tells a story about what it will do – modernise the bureaucracy, close service gaps, and make the state as responsive as the apps already on its citizens’ mobile phones. Pakistan has been hearing a version of this story since the federal cabinet approved the National AI Policy in July 2025, and the story has only grown more confident with time, as officials now speak of a billion-dollar ecosystem and a new seat at an international AI body. What this story does not yet have is a reliable way of checking itself against what citizens actually experience when an AI-assisted government service malfunctions, or when it simply fails to function as it was promised to.
In the realm of political science, this gap between the story a government tells and what the public does with it is called a policy narrative: built, in the Narrative Policy Framework’s own vocabulary, out of a setting, characters, a plot and a moral, and tested for whether it actually shifts what people believe. A 2025 comparative study found that Chinese and American AI policy narratives respond to public attention in strikingly different ways. Researchers have shown how governments strategically construct AI narratives within their own policy documents, how AI itself can now author persuasive policy messaging that measurably increases citizens’ willingness to comply, and how China’s local administrations are using AI to make governance more “agile” – assuming, in every case, that there exists one reasonably coherent government narrative to study in the first place. This method has already travelled well beyond China: a 2026 study by Anne Goldmann and Sandra Plümer traces how the European Parliament itself constructed and revised its AI Act narrative over time, and a separate study of local-government AI adoption finds that public support tracks trust and perceived risk considerably more than it tracks the technology itself.
Pakistan has only just begun measuring the size of the gig economy it is being asked to regulate, let alone measure how fairly it treats workers and consumers
No published study has yet applied this method to Pakistan, and the gap described in the introduction – that a policy which was dithering in February is helping build a billion-dollar ecosystem by October – is exactly the kind of material it would use: a government narrative of steady progress, a public record of stalled delivery, and no systematic attempt as yet to measure how, or whether, the distance between the two is closing. China, for its part, is not done either: the NDRC announced in early November 2026 that it will accelerate work towards comprehensive, unified AI legislation, having previously relied on a patchwork of sector-specific rules. The most significant implication here is that even the state furthest ahead on narrative and measurement still treats its own framework as unfinished business.
Adaptation, Not Imitation
None of this is an argument for importing China’s toolkit wholesale, and the reason is structural rather than cultural. The mechanisms described above – narrative-salience tracking, machine-learning satisfaction models, a quantitative consistency index for digital economy policy documents – all assume two conditions that Pakistan, especially after the stunted devolution of powers under the rubric of the 18th Amendment, simply does not have. The first is a single administrative hierarchy running continuously from the centre down through the provincial and local tiers, which is what lets a researcher treat China’s local government architecture as one reasonably uniform object of study. The second is a single, centrally authored policy narrative, which is what gives narrative-tracking methods something coherent to measure against public attention in the first place. Pakistan’s narrative environment, by contrast, is produced by multiple federal ministries, provincial departments and private actors, with no central authoring or coordinating body stitching them into one story: this is a fact about who produces the narrative, and definitely not a verdict on anyone’s right to say what they choose.
And it must be conceded that this fragmentation looks, on the surface, like a disadvantage; it is certainly why no Pakistani study has managed the kind of clean, single-object narrative analysis that Chinese and European researchers have already published. But the fragmentation is also the actual shape of the problem Pakistani public administration research has to solve: pretending otherwise would only produce answers and policies that quietly assume away the hardest part of it.
Narrating, measuring and regulating are all exercises of discretion over how the state treats citizens, raising questions of accountability, legitimacy and implementation that have no technical solutions
An adapted version of each mechanism would have to track several competing narratives at once rather than one; build federalism-aware, province-differentiated satisfaction models instead of a single national index, given how unevenly provincial administrative capacity is distributed; and, for platform regulation, lean on yet another hybrid arrangement – this time formalising a working relationship with an evaluator that already has standing, rather than attempting to build from scratch a unilateral state apparatus for monitoring, which would ostensibly be staffed by specialists but headed by bureaucratic generalists. Existing work comparing Pakistani e-governance to the Chinese model and comparative institutional frameworks tested across the BRICS economies both point in this direction already, without quite naming it as the research agenda that it actually is. Until someone does the harder, fragmented version of this work, Pakistan will keep narrating, measuring unevenly, and regulating piecemeal – three separate activities that are already beginning to converge into one, as AI-generated policy narratives measurably shape the compliance they are meant to explain.
Narrating, measuring and regulating are all exercises of discretion over how the state treats citizens – questions of accountability, legitimacy and implementation – not technical problems with technical solutions, which is exactly the traditional subject matter of public administration rather than computer science or AI-safety research. An AI or data-science study can only tell you how an algorithm classifies a citizen or predicts their satisfaction, whereas a public-administration framing can tell you whether that classification is legitimate, who is accountable when it’s wrong, and what the state owes the person on the other end of it. This is why the convergence traced herein is a public administration research agenda first, and becomes a technical inquiry only in a secondary, narrower frame.
The Cost of Not Counting
The electronic form that asks twice for the same document, the SMS that explains nothing, the ride booked on hope and faith rather than evidence of experience – these are not really stories about technology. They are about the absence of any mechanism that would tell Pakistan’s own state, let alone its citizens, whether each experiment it has started – AI in administration, digital government, platform regulation – is actually working. China’s researchers did not begin by measuring either; they began, much as Pakistan has, by building first and asking questions later. What changed was that someone eventually built the measuring instruments too, no matter how centralised or incomplete, and treated the gap between the official story and the public’s experience of it as worth studying rather than worth managing away.
That gap is exactly as visible in Pakistan today as it was in China some five to eight years ago, and it remains almost entirely unstudied in Pakistan. Whoever closes it first – in a journal article, a policy paper, or simply by building the index no one has built yet – will not be copying China. They will be doing, for the first time, in Pakistan’s own fractured and plural way, something China’s local governments have already started doing for theirs.
It may be argued that none of this is urgent yet, since no crisis has forced the question. But modernisation and automation do not wait for measurement to catch up: every unaudited algorithm reshaping the gig economy, and every digitised government service rushed online without a way to check it, compounds the same unexamined base Pakistan is already building on. Left unaddressed, that base will not fail quietly – it will fail at scale, in a state and a private sector both of which are moving faster than anyone can track them before they falter, or rein them in before they begin to fail.