A senior researcher at the Lahore University of Management Sciences has published a critical assessment of Pakistan’s first National Artificial Intelligence and Health gathering, warning that the country’s accelerating interest in Artificial Intelligence for healthcare is moving faster than the evidence, ethics, and sustainability frameworks needed to ensure these tools actually improve outcomes for the populations they claim to serve.
Maryam Mustafa, writing for the Mahbub ul Haq Research Centre at LUMS, reflects on a convening held on May 16 at LUMS under the National Artificial Intelligence Hub that brought together government partners, startups, researchers, academics, and donor agencies for fourteen presentations and a private roundtable with provincial health partners from Sindh, Balochistan, Islamabad Capital Territory, and Punjab. She describes the event as a genuine first for the sector but uses it as a foundation for a frank diagnosis of where Pakistan’s Artificial Intelligence and health discourse falls short.
Her central critique is that virtually none of the presentations used their own data to address the most fundamental question facing the sector: does Artificial Intelligence actually work in Pakistan’s localised context, and what is the measurable impact on patients, providers, and the health system? The discourse instead defaulted to features, partnerships, and product pipelines, leaving the evidence base almost entirely absent. She draws a direct comparison to the pattern of pilotitis documented across sub-Saharan Africa, where a churn of small parallel pilots prevents digital health from consolidating into something health systems can actually depend on, and argues Pakistan is well past the point where this conversation should already have been established. She also flags a disconnect from ground realities, noting that most tools presented were designed in English, for smartphones, for literate users, effectively excluding the majority of the women, frontline workers, and patients these products claim to serve.
On ethics, Mustafa describes the silence in the room as the most concerning aspect of the day. Across fourteen presentations, questions of consent, bias, harm, data ownership, and accountability for tool failure were almost entirely absent. She raises what scholars working on Artificial Intelligence in African health systems have termed data colonialism, the extraction of health data from low-income populations to develop and commercialise products that are ultimately owned and valued by institutions in high-income countries, and argues this dynamic must be named directly rather than treated as a benign by-product of well-intentioned innovation. Her sharpest observation comes from her own fieldwork in Karachi, where she documented a community midwife simultaneously running two separate Artificial Intelligence interventions on two devices alongside four physical registers, her patient count rising because word had spread of an ultrasound pilot, her targets set to be reset upward on numbers that will collapse when the pilot ends and the equipment disappears. Mustafa frames this not as an abstract policy risk but as the direct human cost of pilotitis playing out in a single waiting room.
She concludes with a set of urgencies she argues the sector must adopt, including a problem-first framework that asks whether Artificial Intelligence is the right answer before building, mandatory evaluation plans tied to funding and adoption, explicit ethics infrastructure covering consent and data governance for low-literacy populations, and a target that ultimately the government, rather than donors, should own and fund any tools embedded in public health workflows.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
