Page 31 - CMA Journal (Jul-Aug 2026)

F oc us S ec t ion
Reimagining Procurement in Pakistan: Data Analytics,
AI and Human-Centred Decision-Making
Introduction energy-related services.
For instance, there
Obtaining products and services at the best price is not would be a potential
only the goal of procurement; making pro ts, value of Rs. 200 million
managing working capital, operational e ciency, for a hypothetical 2
vendor association, tax compliance, and nancial risk percent e ciency
all are impacted. Pakistan's projected procurement improvement on an
spending demonstrates its potential economic annual procurement
signi cance, and the World Bank has emphasized the expenditure of Rs. 10
global scope of public procurement (World Bank, billion. This computation
2018). Contract management, supplier rivalry, shows the nancial
purchasing e ciency, and fraud prevention can all importance of
bene t greatly from even small improvements. Digital procurement analytics Abbas Ali, ACMA
procurement adapts the basis of judgment. for Karachi's industrial Senior Assistant Manager Finance
Organizations can use transactional data to compare base, but it is not a Lucky Cement Ltd.
prices, spot trends, appraise suppliers, and identify reported savings.
di erences instead of mainly depending on individual
experience and past performance. Therefore, the main Pakistan's Digital Procurement
objective is not whether procurement should go Transformation
digital, but rather how AI and digital systems can With EPADS, Pakistan has entered a new stage of digital
improve decision-making without surrendering procurement. According to PPRA, during
professional judgment, accountability, and FY2024–2025, EPADS 1.0 processed transactions
governance. Instead of presenting original empirical totaling more than Rs. 1.4 trillion, with 51,000 suppliers
research, this paper uses a conceptual and and over 10,000 public-sector organizations
practitioner-oriented approach, synthesizing publicly registered. Additionally, PPRA reported administrative
available regulatory, institutional, and industry data. cost savings of about Rs. 20.52 billion. With supplier
Karachi and Pakistan's Procurement registration, bene cial ownership veri cation,
Context electronic invoicing, digital payments, electronic bid
submission, system-based bid evaluation, and
A critical context for understanding the potential of contract-management capabilities, EPADS 2.0
digital procurement is provided by Karachi. The city broadens this model (PPRA, 2026). By July 2026, 18,909
has a diverse procurement environment surrounding vendors and 1,824 procuring agencies had registered
trade, infrastructure, transport, healthcare, water and with EPADS 2.0; since February 2026, about 7,000
health, construction, logistics, technology, and procurements had been nished. The reported
professional services since it is Pakistan's main integration with FBR, NADRA, SECP, PEC, provincial
commercial and industrial hub. FY2025–2026, the revenue authorities, and DRAP is noteworthy due to
Sindh government announced a growth budget of Rs. the growing connections between procurement data
1.018 trillion, which comprised Rs. 85.5 billion for and corporate information, tax records, regulatory
signi cant Karachi projects under the Local databases, and oversight mechanisms. Thus, in
Government Department (Government of Sindh, addition to traditional purchasing skills, future
2025). Numerous suppliers, contracts, invoices, procurement professionals will need to be pro cient in
payments, and procurement decisions are involved in ERP, taxation, analytics, digital controls, and arti cial
this expense. Analytics can nd variations in prices intelligence.
paid for comparable products, suppliers who regularly
ful ll delivery commitments, suppliers with high From Procurement Data to
rejection rates, underutilized contracts, odd Evidence-Based Decisions
purchasing patterns, and chances for supplier Large amounts of data are produced by digital
consolidation in industries like cement, textiles, and systems, but data by itself cannot produce value. When
FMCG, where Karachi-based manufacturers purchase trustworthy data is transformed into proof to back up a
high-volume raw materials, packaging, freight, and choice, value is created.
ICMA’s Chartered Management Accountant, Jul-Aug 2026 29
Supplier spending, price history, delivery performance, category. Purchase order, supplier invoice, ERP
quality, purchase orders, goods receipts, invoices, and validation, FBR digital invoice, sales tax and WHT
validation, payment approval, and bank payment are
payments can all be examined by procurement
analytics. These domains assist organizations in the representation of the linked process. Payment
delays may result from automated exceptions caused
transitioning from descriptive reporting to the
detection of trends, anomalies, and possible hazards.
The practical questions are simple: Which suppliers get by inaccurate NTN data, tax status, tax codes, or invoice
values. On the other hand, missing transactions, odd
the biggest portion of the budget? Do current prices supplier behavior, inconsistent values, and possible
di er signi cantly from past prices? Which suppliers compliance issues can all be found with the use of
ful ll deadlines? Where are price or quantity integrated data.
mismatches occurring? Are there duplicate or odd Automation, AI and Human
payment patterns? Table 1 summarizes how each Accountability
category of procurement information generates Automation is becoming more and more appropriate
speci c evidence and supports a corresponding for routine tasks like data entry, basic quotation
management action. comparison, purchase order generation, invoice
Procurement Analytics, Accounts matching, duplicate detection, routine reporting, and
Payable and FBR basic supplier classi cation. Instead of just eliminating
jobs, this is better understood as job transformation.
Accounts Payable, Finance, and Procurement shouldn't Future workers might devote more of their time to
function separately. The Purchase-to-Pay cycle strategic sourcing, negotiations, contract
(Requisition, Purchase Order, Good or Service Receipt, management, risk assessment, cost modeling, supplier
Invoice, Tax Validation, and Payment) establishes a development, regulatory interpretation, AI
direct link between nancial controls and commercial monitoring, and exception management rather than
decisions. Unusual payment patterns, duplicate
invoice numbers, incomplete documentation, invoices administrative tasks. Spend analysis, risk management,
without approved purchase orders, invoice quantities supplier identi cation, negotiation optimization,
exceeding received quantities, and invoice prices compliance checks, fraud detection, anomaly
exceeding approved purchase-order prices can all be detection, supplier evaluation, demand forecasting,
agged by analytics. Three-way matching is market-price analysis, and contract review can all be
strengthened and duplicate payments, overpayments, aided by AI (Organization for Economic Co-operation
and processing errors can be decreased by integrating and Development [OECD], 2025). Decisions about
procurement and accounts payable data. An additional procurement, however, are not entirely mathematical.
degree of control is provided by the electronic While a procurement manager is aware of recent
invoicing system of the Federal Board of Revenue nancial, quality, or capacity issues that are not shown
(FBR). According to FBR, noti ed corporate and in the data, a supplier might have a solid track record.
non-corporate registered persons are required to As a result, an AI recommendation might make sense
submit electronic invoices under S.R.O. 709(I)/2025 statistically but be unsuitable for business. Data
dated 22 April 2025 (Federal Board of Revenue, 2025), representativeness and quality are also important
with phased implementation dates speci ed for each factors for AI systems.
Biased historical data may favor certain suppliers or Taxation data, and business history. (2) Integrate tax,
replicate current aws. Additional risks include vendor Enterprise Resource Planning, and procurement
systems to rise transparency and close settlement
lock-in, out-of-date regulations, limited transparency,
and skills gaps. The risk can be stated as follows: Poor gaps. (3) Set human approval boundaries to permit for
the automation of routine, low-risk transactions while
data quality, mismatched systems, and skill shortages;
requiring documented approval for high-value,
wrong decisions and large-scale impact because
automated logic can be applied at scale. Therefore, strategic, or unusual decisions. (4) Create AI
human review, documented approval, and audit governance guidelines that address approved uses,
private information, recommendations made by AI,
controls are crucial for strategic or high-value human review, accountability, and audit trails, and
decisions. AI-assisted decision-making, not AI-owned
decision-making, should be the guiding principle. ensure that signi cant decisions made with AI
Building or Acquiring AI Capability assistance are traceable and recorded. (5) Develop the
skills of procurement professionals in data analytics,
Businesses should carefully consider whether to create Excel, Power BI, ERP, taxation, and AI while keeping an
in-house AI systems or depend on outside vendors. eye on AI performance for bias, accuracy, security, and
Although widely available tools can be reasonable and evolving business conditions and avoiding
applied, there are con dentiality and safety risks when over-reliance on a single technology vendor.
uploading private procurement data. While enterprise
solutions may o er more control, they may also lead to Conclusion
vendor dependence. Customization and control over With the use of FBR's digital invoicing framework, ERP
complex data may be possible with an internal systems, and EPADS, Pakistan is progressing toward a
platform, but infrastructure, cybersecurity, data more integrated procurement environment. The goal
science, and maintenance investments are essential. In of this transition should be to enhance
addition to Pakistan's varying data protection decision-making's quality, speed, transparency, and
regulations, organizations should consider data dependability rather than just replace paper with
software. AI is capable of analyzing vast amounts of
position, cross-border transfer, and concealment data and identifying trends, but it can also make
implications when processing or hosting data outside inaccurate recommendations when the data is biased,
of Pakistan. Keeping AI should not be the only goal.
Control, con dentiality, and accountability should be incomplete, or poorly managed. Automation may
upheld. Whensoever possible, companies should avoid decrease monotonous work while raising the need for
excessive vendor lock-in, maintain control over crucial experts in technology, nance, taxation, procurement,
procurement information, and make sure that and analytics. Future research could monitor these
decisions made with AI support can be reviewed. metrics as the system develops. One limitation of this
Regional Lesson and Future Model paper is that the EPADS adoption gures cited are
self-reported by PPRA at an early stage of nationwide
An appropriate example is a Bangladesh as per world rollout and should be interpreted as directional rather
Bank Bangladesh's electronic government than independently audited outcomes. "Automate the
routine, analyze the data, challenge the AI, and keep
procurement system touched about 80% of its public
procurement expenditures in 2022, up from just 2% in humans accountable for the decision" is the guiding
2017. Furthermore, there were about 7% cost savings principle. This strategy can assist Pakistan in turning
procurement from a compliance-focused function into
compared to manual paper-based procurement
(World Bank, 2022, 2023). The number of registered a strategic ability that improves transparency,
bidders improved while the lead time for procurement strengthens nancial and tax controls, and adds value
reduced. The experience shows that digital to industries and the country's economy.
procurement is more than just an Information
technology project; it calls for formal change, supplier References
participation, dependable data, user instruction, • Federal Board of Revenue. (2025). Digital invoicing: Frequently asked questions
and technical assistance. Government of Pakistan.
governance, and continuing observation. Therefore, • Government of Sindh. (2025). Murad reviews development portfolio 2025–26
human + data + AI + governance should be used to with Rs. 1.018 trillion outlay. Government of Sindh.
describe the upcoming future. Data o ers proof, • Organisation for Economic Co-operation and Development. (2025). Governing
with arti cial intelligence: The state of play and way forward in core government
analytics nds trends, AI predictions and proposes, functions. OECD Publishing.
procurement experts use business decision, nance • Public Procurement Regulatory Authority. (2026). EPADS 2.0: Digital
transformation of public procurement in Pakistan. Government of Pakistan.
and tax verify penalties, and governance creates • World Bank. (2018). Pakistan at 100: Governance and institutions. World Bank.
responsibility. Automation is used to remove • World Bank. (2022). World Bank-funded project drives public procurement
reform in Bangladesh. World Bank.
monotonous, low-value tasks so that professionals can • World Bank. (2023). World Bank helps Bangladesh transform public procurement
focus on making higher-value decisions. into a sustainable online system. World Bank.
Recommendations for Pakistan
The above-mentioned study suggests that Pakistani
organizations and regulators take the following
measures. (1) Improve the quality of obtaining data by
normalizing product classi cations, supplier records,