Let us be honest about something Pakistan’s governance discourse rarely says plainly: political parties do not want local governments to work. This is not a failure of will or capacity. It is a deliberate choice, repeated by every party in every province across every cycle of government, regardless of ideology, coalition or electoral mandate. Local bodies are constituted late, funded poorly, stripped of administrative authority and dissolved early. The pattern is too consistent, too cross-partisan and too durable to be accidental. It is policy by other means. A rival power centre The reason is straightforward. A functioning local government is a rival power centre. It has an electoral base, a budget, a staff and a mandate to deliver services that citizens can see and touch. It can build a political constituency independent of the provincial party machine. For parties whose power rests on controlling the distribution of roads, postings, contracts and patronage from the provincial capital, that is an existential threat not to the state but to them. So they keep the tier alive enough to satisfy constitutional requirements and dead enough to pose no challenge. The local body hence becomes a holding cell for lower-tier party workers, not a governing institution. Waiting for parties to change this calculus is not a strategy. No party in power has ever voluntarily strengthened a tier of government that could constrain it. The few periods of active local government in Pakistan’s history: the 1960s Basic Democracies, the 2001 devolution under Musharraf and the post-18th Amendment councils were all driven by either military governments with no electoral stakes or by court orders that were subsequently ignored by political parties. Democratic parties have never, left to themselves, built strong local governments. Expecting them to do so now is to misread the incentive structure entirely. Meanwhile, the citizen’s need for governance does not wait. Sewage backs up. Schools sit without teachers. Basic registrations, certificates, and permits require travel to a city three hours away. Climate shocks, floods, heat waves and droughts demand local response capacity that no provincial capital can provide at speed. The gap between what people need and what the state delivers grows wider each year and the existing institutional architecture with four massive provinces, a thin federal tier and a local layer that exists mostly on paper cannot close it. The fix If parties will not allow local governments to function, the logic points toward a different structural fix; smaller administrative units. Not smaller local bodies within existing provinces, but the provinces themselves reconfigured or supplemented by administrative divisions drawn on rational, service-delivery grounds. Units small enough that a single administration can actually sense and respond to local conditions. Units where the district health officer knows the district and where the education planner is not managing the scale of a mid-sized European country from a single office building. The critical question is the basis on which these units are drawn. Ethnicity and language are the obvious political demands, but they are the wrong answer. Ethnic administrative boundaries encode group competition into the map permanently. Every resource allocation becomes a proxy for ethnic power. Every appointment is a signal about whose territory this is. Countries that draw administrative lines on ethnic logic spend their governance energy managing the boundary itself rather than delivering services within it. Pakistan, with its layered ethnic, linguistic, sectarian and tribal identities — many of which overlap, shift and are internally contested — cannot afford administrative units that calcify those identities into state structures. The alternative is ruthlessly administrative logic: population thresholds, geographic coherence, economic connectivity, river basin and infrastructure alignment and service delivery catchment areas. A unit that can support a tertiary hospital, a polytechnic institute, a land records office and a local court without borrowing capacity from a neighbouring unit. Boundaries drawn by asking what geography enables efficient governance, not by asking which group lives here. This is not ethnicity-blind; a sensible administration accounts for linguistic diversity in staffing and communication but it refuses to make ethnicity the organising principle of the state’s internal architecture. The tools are available This is where the twenty-first century offers Pakistan something it did not have in earlier rounds of administrative reform: the tools to actually run small units efficiently. Artificial intelligence, digital service delivery and real-time data platforms are collapsing the administrative cost of small-scale governance. A local administration no longer needs to physically house every function. AI-powered triage systems can route citizen complaints, allocate maintenance crews and flag anomalies in service delivery faster and more accurately than a provincial bureaucracy managing 10 million people from a single dashboard. Land records, business registrations, birth and death certificates and public grievance systems can all run on shared digital infrastructure, with the local unit as a point of contact rather than the processor. In an AI-assisted governance model, the case for large administrative units historically justified by economies of scale in bureaucratic capacity largely disappears. What you need at scale is a shared technology layer, shared standards and interoperable data systems. What you need locally is a small, elected or appointed administration with genuine authority, a budget tied to local revenue and population and accountability to the people who live there. The AI layer handles the complexity that used to require size. The local layer handles the judgment, the presence and the trust. Several countries are already moving in this direction. Estonia’s entire public administration operates at a scale: 1.3 million people that is comparable to a mid-sized Pakistani district and its digital governance stack is the backbone of service delivery rather than a supplement to bureaucracy. Rwanda reorganised its districts in 2006 on service-delivery logic, not ethnic logic after the catastrophic lesson of what ethnic administrative boundaries produce. South Korea’s smart city initiatives embed real-time sensing and AI-driven resource allocation at the municipal level. The direction is consistent: governance is becoming more local, more data-driven and less dependent on administrative mass. Pakistan’s current map is a colonial inheritance optimised for revenue extraction and security control, not for citizen service. The four-province structure, including Gilgit Baltistan and Azad Jammu and Kashmir, whatever its historical justification, was never designed around the question of what unit of government could best respond to a family that needs a functioning school, a clean water connection and a place to file a legal complaint. Answering that question now with the tools available in 2026 points toward smaller units, administrative boundaries and a digital backbone that gives those units the capacity that size used to provide. The parties will resist. They always do. But the argument for smaller, rationally drawn administrative units does not require party consent; it requires a constitutional conversation about what the state is for. Pakistan’s citizens have been patient with an architecture that was never built for them. The AI era doesn’t just make better governance possible. It makes the old excuses for not delivering it obsolete.