Page 46 - CMA Journal (Jul-Aug 2026)

F oc us S ec t ion

The risk becomes more signi cant in procurement Similarly, Brazilian construction company Odebrecht Connecting Data Across Systems of analysing each additional transaction can fall For countries where internal audit teams are controls such as segregation of duties, audits and
when individuals have control over substantial was involved in corruption across multiple countries substantially. consistently outnumbered by procurement volumes, supplier checks remain important, but technology can
nancial resources and considerable decision-making and ultimately paid substantial penalties and Fraud schemes can also exploit gaps between di erent In some cases, organisations may achieve a return on automated tools are increasingly becoming a necessity extend their reach by analysing transactions at scale,
authority. settlements. The Petrobras “Car Wash” scandal also organisational systems rather than weaknesses within investment within six to twelve months when avoided rather than simply a modernisation initiative. identifying unusual patterns and connecting
resulted in signi cant nancial losses associated with a single system. Cross-system AI can therefore provide Technology can expand the coverage of audit and information across multiple systems.
In developing economies, reliance on manual additional value by correlating information from email, losses and sta time savings are considered. However,
processes can further increase the opportunity for overpriced contracts. These cases demonstrate how enterprise resource planning (ERP) systems and successful implementation depends heavily on proper monitoring functions without requiring organisations The objective is not to eliminate human involvement
fraud. Physical les can be altered, removed or procurement fraud and corruption can become deeply integration and con guration. Poorly tuned AI systems to increase their audit workforce proportionately. from procurement oversight but to enable
embedded in business operations when fraudulent
Due to limited resources, it is not possible for every companies established destroyed, while fragmented records may make it practices remain undetected for extended periods. Continuous Transaction Monitoring: operating on inaccurate or incomplete vendor data The Future of Anti-Fraud Procurement professionals to focus their expertise where it matters
organisation to produce everything independently. for such purposes may, di cult to establish a complete audit trail. Every Transaction, not a Sample can generate excessive false alerts and may ultimately most. AI, machine learning, continuous monitoring,
Resource constraints have therefore compelled upon investigation, be What Traditional Controls Missed What Technology Is Doing Di erently AI / ML cost more than the losses they prevent. Despite the potential of these technologies, their blockchain and stronger vendor-data controls can
Anomaly
organisations at the national, regional and global linked to an employee’s One of the most signi cant changes in modern Transaction Scoring Detected? Flagged for Investigator Why Technology Matters More in adoption should not be interpreted as an argument for provide earlier warnings and broader visibility, while
Human
Occurs
Decision
levels to work together to ful l one another’s needs. address, bank account or For decades, the primary tools for preventing and procurement is the shift towards continuous Engine Review replacing human judgment with algorithms. accountants, auditors and procurement professionals

This may involve the supply of raw materials, high-tech other personal interests. detecting procurement fraud were largely manual and transaction monitoring. Instead of reviewing a sample Instant, automated — not an annual audit cycle Developing Economies Presenting technology in such terms would overlook remain responsible for interpreting those signals and
equipment and machinery, as well as services such as Other forms of periodic. These included segregation of duties, of transactions periodically, automated systems can Figure 3: Continuous monitoring reviews every transaction the moment The case for adopting anti-fraud technology becomes important limitations associated with automated taking appropriate action.
construction and consultancy. Procurement is the procurement fraud post-transaction internal audits, policy checks, assess transactions as they occur and identify it happens, rather than a periodic sample particularly compelling in developing economies decision-making. Ultimately, anti-fraud technology is most e ective
process of identifying needs, searching for suppliers or include invoice fraud reference checks and supplier blacklists. Such controls potentially unusual activity. For example, a system can where audit capacity may be limited relative to Research and practice concerning trustworthy AI in when it is supported by reliable data, sound
vendors, and negotiating the prices and costs of goods and overbilling, remain valuable and have successfully identi ed ag duplicate payments as soon as they are recorded. payment platforms. This can help identify patterns that procurement volumes. public procurement continue to highlight the governance and professional judgment. The real
or services. While de nitions may vary slightly across undisclosed con icts of numerous cases of procurement fraud. However, their may remain invisible when each system is analysed importance of e ective governance. If an AI model is opportunity is therefore not simply to automate
sources, an e ective procurement process generally interest, splitting e ectiveness is limited by their reliance on sampling Arti cial intelligence and machine learning can take separately. South Asia provides several examples of digital deployed without appropriate oversight, it may existing controls, but to create a procurement
incorporates a number of core operational steps. purchases into smaller and periodic review. this capability further by identifying deviations from The Impact of Anti-Fraud Technologies procurement at di erent stages of development. incorrectly ag a legitimate supplier as suspicious environment in which unusual behaviour is identi ed
India’s Government e-Marketplace (GeM), launched in
For any organisation, procurement represents one of transactions to Sampling can detect only those transactions selected normal purchasing patterns across vendors, categories 2016, represents a relatively mature example. without providing a su ciently transparent basis for earlier, investigated more e ectively and becomes
the areas through which substantial amounts of circumvent approval for examination. A fraudulent scheme spread across and transactions. The systems can identify behaviours According to Feedzai/Orbograph industry survey data, According to the available data, it connects more than the decision. Such an outcome can cause signi cant increasingly di cult to conceal.
money can leave the organisation without immediate thresholds, and change-order abuse. In the latter case, hundreds or thousands of transactions may continue such as a vendor repeatedly submitting bids that are close to 40% of nancial institutions that had adopted 1.6 million buyers with over 2.3 million sellers. An harm to businesses and undermine con dence in the References
visibility. Audit professionals frequently identify a supplier may submit a low winning bid and for years without appearing in a sampled transaction. just high enough to lose while remaining part of a AI-based fraud detection reported reductions in losses independent World Bank evaluation found that buyers procurement process.
procurement as a high-risk area, not necessarily subsequently increase the contract value through Furthermore, when controls are applied primarily recurring bidding pattern, or purchases being of between 40% and 60%. Ukraine’s ProZorro achieved median price savings of approximately An AI-based system should therefore be su ciently • Association of Certi ed Fraud Examiners, Report to the Nations — the
fraud triangle framework of opportunity, pressure, and rationalisation is a
because procurement personnel are dishonest, but variations or additional work that was not subject to during an annual audit, the funds may already have deliberately divided to remain below approval procurement system, which combines transparency 9.75%. At the same time, India’s CAG has highlighted explainable for auditors and procurement long-standing staple of fraud examination literature.
because the process contains numerous discretionary the original competitive process. In practice, been transferred and the supporting documentation thresholds. mechanisms, blockchain-based record-keeping and the need for stronger benchmarking of GeM’s savings professionals to understand why a particular • U.S. Securities and Exchange Commission, “SEC Charges Siemens AG for
points at which decisions can be manipulated without procurement fraud schemes may involve a may have been altered or organised to conceal the Benford’s Law can also be used as an analytical AI-supported agging, has been credited with methodology, demonstrating that digitisation alone transaction or supplier has been agged. The Engaging in Worldwide Bribery,” press release 2008-294, December 15,
2008. sec.gov/news/press/2008/2008-294.htm
detection for extended periods. Procurement fraud combination of two or more of these methods technique to identify unusual distributions of leading generating substantial savings. Similarly, India’s does not automatically guarantee e ective oversight. responsibility of accountants and auditors to exercise • U.S. Department of Justice, “Siemens AG and Three Subsidiaries Plead
occurs when an individual deliberately manipulates operating simultaneously. underlying activity. digits in nancial data. When combined with modern Comptroller and Auditor General (CAG) announced in professional scepticism does not disappear simply Guilty to Foreign Corrupt Practices Act Violations,” press release, December
sourcing, tendering, contracting or payment processes The nancial consequences of undetected anomaly-detection techniques and machine-learning 2026 that it was using AI tools and advanced data Bangladesh’s e-GP system has been operational since because an algorithm makes the initial assessment. 15, 2008. justice.gov/opa/pr/2008/December/08-crm-1105.html
to obtain nancial or other bene ts dishonestly. The procurement-related corruption can be substantial. models, such methods can provide an additional layer analytics to scan public tenders, identify pricing 2011 and had moved most large tenders online by Instead, their role expands to include reviewing the • Harvard Law School Forum on Corporate Governance, “2016 Year-End
critical distinction between an error and procurement Between 2001 and 2007, Siemens paid more than $1.4 of analysis for identifying invoice batches or nancial 2023. Pakistan has also expanded digital procurement model, its thresholds, the quality of its underlying data FCPA Update,” January 19, 2017 (Odebrecht/Braskem settlement detail);
Willkie Compliance Concourse, “Odebrecht Monitorship Comes to a Close,”
fraud is intent: fraud involves knowingly circumventing billion in bribes to secure contracts ranging from records that may warrant further investigation. The Measurable Impact of Anti-Fraud Technology through the Public Procurement Regulatory and the processes through which agged suppliers or November 2020 ( nal $2.6 billion penalty).
established rules or controls for an improper purpose. national identi cation systems in Argentina to metro 40–60% Authority’s Electronic Procurement System (EPADS). transactions are investigated. • Mongabay, “Brazil's Lava Jato Investigation: The Biggest Corruption
Scandal of the Last Decade,” March 2025; Petrobras Form 20-F lings, U.S.
6-12
~$6B

How Procurement Fraud Happens lines in Venezuela. Following its settlement with US Strengthening Vendor Master Data Reduction in Saved by Months 100% During FY2024–25, EPADS reportedly processed Procurement fraud often thrives on transactions that Securities and Exchange Commission (sec.gov/cgi-bin/browse-edgar).

and German authorities in December 2008, Siemens Vendor master data is another important area where fraud losses Ukraine’s Typical Of transactions 526,239 transactions worth Rs. 1.408 trillion and appear routine, insigni cant or anonymous and • AllAboutAI, “AI Fraud Detection Statistics 2026,” citing Feedzai/Orbograph
reviewed under
ProZorro
Procurement fraud generally occurs through several paid approximately $1.6 billion in nes and technology can strengthen procurement controls. reported by platform since payback period monitoring, vs. generated administrative cost savings of Rs. 20.52 therefore receive limited scrutiny. Continuous industry survey data. allaboutai.com/resources/ai-statistics/
continuous
institutions
for AI fraud-
detection
recurring patterns, with bid rigging among the most disgorgement, making it one of the largest Foreign Automated systems can continuously screen vendors after adopting Oct. 2017 platforms, loss a small audit billion. The system has also been connected with monitoring and pattern detection are steadily ai-fraud-detection
through
AI-based
sample
common. In bid rigging, competing vendors collude in Corrupt Practices Act sanctions on record. against relevant watchlists and other control detection⁶ transparency + avoidance + historically institutions including FBR, NADRA, SECP and PEC, with reducing the opportunity for such activity to remain • Harvard Kennedy School, Mossavar-Rahmani Center for Business and
labor saved
AI flagging⁷
Government, “Overcoming Corruption and War: Lessons from Ukraine's

advance to determine who will win a tender. databases. Changes to vendor bank details can be oversight access for bodies such as the Auditor hidden. However, organisations that derive the ProZorro,” 2022; U.S. Strategy on Countering Corruption, December 2021.
Participants may take turns submitting losing bids so subjected to dual approval, while system-based access Figures are directional, drawn from multiple industry sources — see article text for context General, NAB and the Competition Commission. greatest value from these technologies are unlikely to hks.harvard.edu/centers/mrcbg
Figure 4: Headline figures on the impact of anti-fraud technology, drawn
that the group can maintain arti cially high prices controls can restrict who is authorised to create or from multiple industry and public-sector sources. be those that simply invest in the most sophisticated • World Bank, “How Government e-Marketplace Is Revolutionizing
Procurement in India,” World Bank Blogs; Press Information Bureau,
while preserving the appearance of competition. South Asia’s Procurement Build-Out Government of India, GeM assessment note, 2022.
modify vendor records and payment information. anomalies and detect potential bid rigging. These platforms. Rather, the strongest results come from blogs.worldbank.org/en/governance
Another common form is kickbacks and bribery. In Why Procurement Fraud Continues to developments indicate that advanced analytical tools INDIA BANGLADESH PAKISTAN organisations that combine technology with capable • Public Procurement Regulatory Authority (PPRA), Pakistan, o cial

such cases, an o cial directs a contract towards a Blockchain and Record Integrity are increasingly being incorporated into public-sector Government e-Marketplace e-Government Procurement Disposal System (EPADS) professionals who are willing to challenge anomalies announcements on EPADS FY 2024–25 performance and administrative

e-Pak Acquisitions &
(e-GP)

favoured vendor in exchange for a personal bene t. Occur Blockchain technology addresses procurement risks in audit and procurement oversight. Launched 2016 Launched 2011 Launched 2023 and ask the fundamental question that has always savings. ppra.gov.pk/announcements

(GeM)

Such bene ts may be disguised as consulting fees, The recurrence of these fraud patterns across countries a di erent way. Once a record is entered into a properly The initial cost of technology can be higher than that of 1.6M+ government $3B+ annual Rs. 1.4 trillion been central to good auditing: Why should I believe
employment opportunities or arrangements involving and industries can be understood through the designed shared ledger, it becomes signi cantly more traditional manual controls. Manual controls may buyers procurement spend processed, FY 2024- this?
25
covered
relatives or associates. well-established concept of the fraud triangle: di cult to alter it retrospectively without detection. largely be incorporated into existing sta 2.3M+ registered Most large tenders 526,000+ Conclusion
sellers
transactions, FY
online by 2023
A third form involves ctitious vendors. Under this opportunity, pressure and rationalisation. Individuals Such controls can reduce opportunities for fraudulent responsibilities, whereas automated systems require ~9.75% median price Reduced tender- 2024-25
savings (World Bank
evaluation)
board political
Rs. 20.52B admin
scheme, an organisation is invoiced for goods or may rationalise fraudulent behaviour by convincing backdating or unauthorised modi cation of investment in software licences, ERP integration, data interference cost savings, FY The transformation of procurement through data and
2024-25

services that were never supplied or performed. Shell themselves that such conduct is acceptable because procurement records. cleansing and implementation. However, once Figure 5: Three South Asian e-procurement platforms, at three different automation represents a signi cant shift from periodic
others are doing the same. automated systems are operational, the marginal cost stages of maturity. detection towards continuous prevention. Traditional
44 ICMA’s Chartered Management Accountant, Jul-Aug 2026