Pakistan's AI moment: Rs9bn prescription for a structural problem

Pakistan’s IT exports are growing fast, with $3.39 billion in the first nine months of FY 2025–26, a 23.7% year-on-year surge that puts the sector on track for roughly $4.5 billion by year-end — a record. Yet beneath this momentum sits an uncomfortable reality: the growth is happening in volume, not value.

The FY2025–26 budget allocated only Rs4.8 billion to Science & Technology and Rs13.5 billion to the Ministry of IT and Telecom; combined, less than 0.11% of the total Rs17.6 trillion outlay.

The gap between stated ambition and fiscal commitment remains the defining structural risk to Pakistan’s digital growth story.

Pakistan produces more than 75,000 IT graduates annually. Yet only a fraction enters the formal technology workforce, and fewer than 10% of active IT professionals possess applied artificial intelligence (AI) skills.

The reason is structural: university curricula are misaligned with industry demand, and there is no credentialing system connecting AI skill acquisition to employment outcomes.

The Access Partnership estimates that narrowing Pakistan’s digital skills gap could add Rs2.8 trillion to annual gross domestic product (GDP) by 2030. Google’s economic impact research shows AI-related tools already added Rs3.9 trillion in economic benefits in 2023; a 222% increase since 2020. These are not marginal numbers.

The GCC market alone spends over $12 billion annually on AI consulting and implementation services. Pakistan captures less than 1% of that; not because of talent deficits, but because of credibility deficits: no recognised AI training credentials, no institutional track record, and no government-backed certification framework to signal quality to international buyers.

Major IT export firms report that up to 30% of their solutions now incorporate AI components, largely driven by international client demand rather than domestic policy incentives. Pakistani small and medium enterprises (SMEs) are beginning to adopt AI-powered tools, but without structured training or change, management support, adoption remains surface-level.

Pakistan is about to conduct its first-ever IT census; a revealing acknowledgement that the government has been setting export targets for a sector it cannot yet fully measure. Policy cannot be evidence-based if the evidence base does not exist.

R1 · National Applied AI Skilling Fund

Pakistan’s National AI Policy 2025 targets 3,000 AI scholarships annually and the establishment of AI Centres of Excellence in Karachi, Lahore, and Islamabad. These are the right ambitions. They require proportionate funding.

A critical design principle: the fund must prioritise depth over volume. Pakistan already has surface-level AI awareness programmes. What the economy needs are practitioners who can architect and deploy; not professionals who have completed a 4-hour online course.

Accordingly, a two-tier structure with explicit quality safeguards can be recommended:

Tier A; advanced practitioner track (Rs2 billion)

The government should target 20,000 high-tier certifications at approximately Rs100,000 per head. This is deliberately ambitious but not unrealistic: Malaysia’s equivalent programme achieved 18,000 advanced AI certifications in 18 months with comparable per-head investment.

Delivery through private sector partners selected via open, competitive PPRA tendering. Eligibility criteria: prior enterprise AI delivery experience, documented curriculum aligned with PACS Tier 2+ standards, and mandatory outcome reporting (employment rate, project completion, industry verification) as a funding condition.

Note on absorption risk: industry absorption of 20,000 advanced practitioners will require parallel enterprise demand stimulation; NAAF (R4) is the demand-side complement.

Tier B; foundation AI literacy track (Rs1.5 billion):

Integration of AI modules into HEC-affiliated institutions; retrofitting existing CS, business, and engineering curricula, not building parallel programmes.

Target 80,000 foundational certifications across universities and TVETs. Per-head cost approximately Rs18,750, consistent with HEC-funded module delivery norms.

Women in AI initiative (Rs1 billion)

Ring-fenced within Tier A and Tier B allocations; not a standalone stream. Integration design: at least 40% of all Tier A PPRA contracts must demonstrate a women’s participation plan with measurable targets, verified at mid-term review. Tier B institutions with below-30% female enrollment receive conditional funding with improvement milestones. This prevents tokenism and ties funding to outcomes.

Ecosystem infrastructure (Rs0.5 billion)

Shared AI compute credits, open datasets, and curriculum standards; public goods reducing delivery costs for all PPRA-approved providers. Managed by PSEB as a neutral infrastructure layer.

R2 · Tax-Neutral Sandbox

Pakistan’s AI startup ecosystem is in its formation stage. Any fiscal proposal must be designed within Pakistan’s actual constraints: FBR’s mandate to widen the tax net, eliminate exemptions, and meet IMF structural benchmarks. A flat tax holiday is not viable. It would not survive FBR review.

We propose a Tax-Neutral Sandbox; a deferred, performance-linked framework that costs the exchequer nothing upfront:

Deferred tax model for registered AI firms

Firms registered under a new PSEB ‘Applied AI Enterprise’ category defer corporate income tax for three years. Deferred amounts are repaid in years 4 and 5, with a 10% repayment premium waived only if net-new SBP-verified foreign exchange remittances exceed a defined annual threshold. Revenue-neutral over the cycle. Legal template: Pakistan’s existing Special TechnologyZones (STZ) framework provides an administrative precedent for deferred fiscal instruments linked to export performance.

Tiered withholding tax linked to SBP inflows

The 5% withholding tax on IT service exports is restructured as a tiered rate; 0% for firms with verified SBP remittances above Rs10 million annually; 2.5% for firms above Rs2 million; 5% for all others. Tax benefit is earned by bringing verified dollars into Pakistan; not awarded upfront.

25% tax credit on documented AI training investment

Any business; across all sectors; that documents structured AI skills training expenditure through a PSEB-registered provider receives a 25% tax credit. This simultaneously expands FBR’s documented expenditure base and incentivises enterprise adoption.

Simplified GST category for AI EdTech

A clearly defined GST classification for cohort-based AI education businesses, removing current regulatory ambiguity and reducing compliance costs.

R3 · Pakistan AI Certification Standard (PACS)

Pakistan cannot capture high-value AI consulting contracts in the GCC, Malaysia, or Western markets without a nationally recognised AI certification standard. Individual professionals and firms currently have no credential to show international clients that signals verified AI capability; comparable to a Salesforce certification, an AWS Solutions Architect badge, or India’s NASSCOM AI certification.

We recommend PSEB develop and administer PACS, with a critical design principle: do not build from scratch. Adapt and align with existing international standards; IEEE’s AI competency frameworks, ISO/IEC 42001 (AI Management Systems), and GCC-specific procurement credential requirements. Building on established standards accelerates GCC market recognition by 18–24 months compared to a bespoke national framework and reduces development costs significantly.

R4 · National AI Adoption Fund (NAAF)

Digital transformation could generate Rs9.7 trillion ($34.9 billion) in value for Pakistan’s economy by 2030. The majority of that value sits in sectors; manufacturing, retail, healthcare, logistics; where SMEs dominate. Without an adoption fund, that value remains theoretical.

We recommend a Rs3 billion National AI Adoption Fund structured as a matching-grant mechanism:

• Eligible SMEs (registered, tax-compliant, under 500 employees) apply for grants covering 40% of documented AI implementation costs; up to a maximum of Rs 5M per firm. Implementation must be carried out through a PSEB-registered AI firm or certified AI implementation partner.
• Verification of ‘documented AI implementation costs’: Eligible costs are defined narrowly; software licences, integration services, staff training from PSEB-registered providers, and hardware directly attributable to AI deployment. All claims require third party accountant sign-off and PSEB spot audits. This prevents leakage and ensures accountability.
• Productivity metrics: Each grant recipient must submit a baseline assessment before implementation and a verified outcome report at 12 months; covering at least two of: cost reduction (%), revenue growth (%), staff time saved per week, error rate reduction, or customer response time improvement. Independent verification by a PSEB-appointed evaluator.
• Sector priority weighting: manufacturing (30%), healthcare (20%), financial services (20%), agriculture-tech (15%), retail/logistics (15%).
• Conflict of interest management: Grant disbursement decisions are made by an independent technical committee; not PSEB alone; comprising MoITT, SBP, a P@SHA representative, and two independent industry experts. Densight Labs and any other firms with staff on the committee are recused from decisions on grants where they are the implementation partner.

All four recommendations are consolidated under a single AI Delivery Unit (AI-DU) housed under the Prime Minister’s Office; preventing the inter-ministerial coordination failure that historically grounds good policy. The AI-DU model has legal precedent in Pakistan: the CPEC Authority and STZA demonstrate that a narrow-mandate, high authority coordination vehicle can cut across ministries without requiring institutional redesign.

A total of Rs9 billion is 0.051% of Pakistan’s FY 2025–26 total outlay of Rs17.6 trillion. Against an addressable digital transformation value of Rs9.7 trillion by 2030, it may represent the highest-leverage line item available to any ministry this fiscal cycle.