Skip to content
AIzaadi
اردوStart free

BISP and NADRA: AI case studies that already shape Pakistan

BISP اور NADRA: AI کے کیس جو پاکستان کو پہلے ہی بدل رہے ہیں

22 min read

Three ways to see it

  1. BISP uses a Proxy Means Test based household survey to identify eligibility. The PMT score is itself a statistical model trained on socio-economic variables. Over time the survey has been re-run and updated. AI enters when teams use machine learning to detect ghost beneficiaries, validate biometric attendance, and route grievances. Each of these tasks must be done without further excluding women, persons with disabilities, or households in remote districts.

  2. Lessons from BISP for AI projects. One: the dataset's age matters. A household that was poor in 2016 may not be poor today. Models trained on stale data freeze old reality into new decisions. Two: exclusion errors and inclusion errors are not symmetric. A wrongly excluded poor family suffers more than a wrongly included middle income family receives. Set the threshold accordingly. Three: appeal channels must work in the village, not only in district offices.

  3. Project examples to study. Ehsaas Kafaalat used NADRA verification and SMS rails to reach beneficiaries during COVID. Sehat Card uses NADRA based identification at private hospitals. The Pakistan Citizen Portal aggregates citizen complaints across federal and provincial entities; ML helps with routing and duplicate detection. The Asaan Mobile Account, mentioned in earlier lessons, sat on top of NADRA verification. Each of these systems answers to a different ministry or regulator, but they share an identity rail.

Quick check

Quick check: what makes modern AI different from a rule-based program?

The why-tree

NADRA runs facial verification, fingerprint matching, and identity linking at national scale. AI helps spot duplicate identities, detect document tampering, and authenticate biometrics. The flip side is well documented: false rejects for elderly fingerprints, women whose biometrics have changed after work or chemotherapy, and rural users with damaged finger ridges. NADRA has run programs to handle exceptions, but each exception path adds time and cost to citizens who can afford it least.

Try this with Claude

Reflective exercise. Pick a public AI project in your sector. Write one paragraph on how it might wrongly exclude a vulnerable group. Write a second paragraph on the smallest change that would fix it without breaking the larger system. Bring both paragraphs to your next project meeting. The discipline of asking who could be hurt is what separates a competent civil servant from a brilliant one.
Previous lesson