Inventory Operations, Accuracy and Control

Introduction

Inventory management is not completed when materials are purchased or when stock appears in an enterprise resource planning (ERP) system. Value is created only when inventory is received correctly, stored safely, identified accurately, picked efficiently, issued to the appropriate user, returned when necessary, counted systematically, and reconciled to reliable records. These activities constitute inventory operations and control.

Operational weakness produces more than inaccurate balances. It causes stockouts despite apparently adequate stock, excess purchasing, production interruptions, customer dissatisfaction, financial misstatement, safety risks, and opportunities for fraud. Inventory accuracy is therefore both an operational capability and a governance requirement.

This chapter examines the complete physical and information flow of inventory. It focuses on process design, internal control, cycle counting, investigation of discrepancies, and audit-ready documentation. The discussion is relevant to manufacturing, distribution, retail, healthcare, construction, and service organizations.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Design controlled receiving, storage, picking, issue, return, and reconciliation processes.
2. Explain how master data, transaction discipline, layout, and human behavior influence inventory accuracy.
3. Calculate and interpret inventory accuracy, shrinkage, and reconciliation measures.
4. Design risk-based cycle-counting programs and investigate record errors.
5. Establish segregation of duties, authorization, evidence, and audit trails for inventory operations.
6. Evaluate the use of barcodes, RFID, warehouse-management systems, analytics, and artificial intelligence in inventory control.

1. Inventory Operations as an Integrated Control System

Inventory operations convert a physical flow into an accountable information flow. Each movement should answer five questions:

• What item moved?
• How much moved?
• From where and to where?
• When did it move?
• Who authorized and recorded it?

A useful control principle is record at the point of activity. Transactions should be captured when goods are received, moved, picked, issued, returned, or adjusted—not reconstructed later from memory or paperwork. Delayed recording creates timing differences and encourages informal workarounds.

The operational sequence is usually:

receipt → inspection → identification → put-away → storage → replenishment → picking → issue/dispatch → return or adjustment → counting → reconciliation

This is not a purely linear process. Returns may re-enter inspection; damaged goods may move to quarantine; and cycle-count findings may initiate investigation rather than immediate adjustment.

1.1 Inventory control objectives

An effective system seeks to achieve the following objectives:

These objectives resemble financial-control assertions, but they must be implemented through physical procedures, system controls, supervision, and evidence.

2. Receiving: Establishing the First Reliable Record

Receiving is the point at which responsibility commonly transfers from a supplier or carrier to the organization. Errors at receipt propagate through every later process, so receiving should never be treated as simple unloading.

2.1 Pre-receipt preparation

Before arrival, the receiving function should have access to a purchase order, advance shipping notice, delivery appointment, or other authorized reference. The item master should contain a unique stock-keeping unit (SKU), description, unit of measure, conversion factors, storage requirements, and, where relevant, lot, serial, expiry, or hazardous-material information.

The receiving area should distinguish between:

• expected and unexpected deliveries;
• accepted, awaiting inspection, rejected, damaged, and quarantined goods;
• owned inventory, customer-owned inventory, and supplier consignment stock.

2.2 Receipt procedure

A controlled receipt normally includes these steps:

1. Verify carrier, supplier, purchase order, seal, and shipping documents.
2. Inspect packaging for visible damage or tampering.
3. Count cartons, pallets, or handling units before opening where appropriate.
4. Open and count contents using the authorized unit of measure.
5. Check item identity, specifications, lot or serial numbers, and expiry dates.
6. Record shortages, overages, substitutions, and damage separately.
7. Apply identification labels and status labels.
8. Move goods to inspection, quarantine, or approved storage.
9. Complete the goods-receipt transaction only after the required checks.
10. File evidence, including photographs or discrepancy reports when needed.

Blind receiving—where the receiver does not see the expected quantity—can reduce confirmation bias, but it may increase processing time. It is most useful for high-value, theft-prone, or frequently miscounted items and should be supported by clear item identification.

2.3 Quantity and quality acceptance

Quantity acceptance and quality acceptance are distinct decisions. A receiver may confirm that 100 units arrived without confirming that the units meet engineering or quality requirements. The system should therefore support statuses such as “received pending inspection” and “released to stock.”

For controlled or regulated goods, records may need to include supplier batch, certificate of analysis, temperature history, chain of custody, or expiry date. In such environments, the receiving process is part of product safety and regulatory compliance, not merely inventory administration.

3. Storage, Identification, and Location Control

Storage design affects travel time, damage, misplacement, counting reliability, and security. Good storage is based on planned locations rather than personal memory.

3.1 Location architecture

Each location should have a unique code, physically displayed and logically represented in the warehouse system. A location code may identify building, aisle, bay, level, and position. Location labels should be readable from the normal approach direction and should not be reused ambiguously.

A location-control policy should specify whether mixed SKUs, mixed lots, or mixed ownership may share a location. Mixed storage can increase space utilization but raises identification and counting risk. If mixed storage is permitted, each quantity must be physically segregated or electronically traceable.

3.2 Slotting principles

Slotting assigns items to locations using criteria such as:

• demand frequency and volume;
• cube, weight, and handling characteristics;
• required picking method;
• compatibility and hazard classification;
• temperature, humidity, security, and expiry requirements;
• replenishment frequency and ergonomics.

Fast-moving items are often positioned near dispatch or at ergonomic heights, while slow-moving or reserve stock occupies less accessible locations. However, velocity alone should not determine placement: hazardous, fragile, high-value, or temperature-sensitive items require priority controls.

3.3 Identification and unit of measure

Many inventory errors are caused by unit-of-measure confusion—for example, receiving cases but issuing individual units, or recording kilograms while planning in tonnes. Every transaction should identify both the quantity and the unit. Conversion factors must be controlled in the item master; users should not create informal conversions in spreadsheets.

A practical label includes SKU, description, quantity, unit, lot or serial number, status, receipt date, and expiry where relevant. Labels should be replaced when damaged rather than overwritten in a way that obscures history.

3.4 Stock status and physical protection

Usable, damaged, expired, returned, rejected, consigned, and customer-owned inventory should not be indistinguishable. Physical segregation is preferable, but system status is also essential. Security measures may include restricted access, cages, CCTV, seal control, key logs, and dual authorization for high-value items.

4. Picking, Issue, and Dispatch

Picking converts an authorized demand signal into a physical selection. The demand signal may be a sales order, production order, maintenance request, transfer order, or approved requisition.

4.1 Picking controls

A picking instruction should identify SKU, description, quantity, unit, source location, lot or serial requirements, destination, and priority. Common methods include single-order picking, batch picking, zone picking, and wave picking. The appropriate method depends on order profile, facility layout, labor, and system capability.

Accuracy controls include:

• barcode or RFID confirmation of item and location;
• scan-based quantity confirmation where feasible;
• independent verification for high-value or safety-critical items;
• system enforcement of lot, serial, and expiry rules;
• exception reporting for substitutions and short picks;
• physical separation of picked orders awaiting dispatch.

4.2 FIFO and FEFO

First-in, first-out (FIFO) issues the oldest receipt first. First-expiry, first-out (FEFO) issues the stock with the earliest expiry date first. FIFO is not always equivalent to FEFO: a later receipt may expire earlier because it had a shorter shelf life at receipt. FEFO is therefore normally required for pharmaceuticals, food, chemicals, and other expiry-sensitive goods.

4.3 Issue to internal users

An issue should be linked to an authorized cost centre, work order, production order, project, or customer order. The issuing employee should record the actual quantity released, not merely the quantity requested. Partial issues, substitutions, and backorders require explicit status codes.

For production materials, backflushing may automatically issue components based on completed output. It can be efficient, but it is vulnerable to inaccurate bills of material, scrap, rework, and timing differences. Physical verification and periodic variance analysis remain necessary.

4.4 Dispatch control

Before dispatch, personnel should reconcile the picked quantity to the packing list and shipping document. The dispatch record should establish custody transfer and preserve evidence of what was shipped, to whom, when, and under which order. Proof of delivery should be connected to the shipment record rather than maintained as an isolated document.

5. Returns, Damage, and Adjustments

Returns are a major source of hidden inventory error because returned goods may be placed back into available stock without inspection.

A controlled return process classifies goods as:

• saleable and suitable for immediate restocking;
• saleable after inspection or repackaging;
• repairable or awaiting technical assessment;
• damaged, expired, or obsolete;
• supplier-return or customer-return stock;
• scrap or disposal stock.

The return document should reference the original issue, order, or shipment. Quantity, condition, lot or serial number, and disposition should be recorded. A return should not automatically increase available stock; it should first enter an appropriate status.

Adjustments are corrections to the record rather than ordinary movements. Every adjustment should have a reason code, authorization threshold, supporting evidence, date, user identity, and reference to the investigation. Frequent manual adjustments are a symptom requiring root-cause analysis, not evidence of good control.

6. Inventory Accuracy and Reconciliation Measures

Inventory accuracy has several valid definitions. A single aggregate percentage can conceal serious errors, so organizations should report both record-level and value-level measures.

6.1 Record accuracy

For a population of counted stock-keeping units or locations, a simple count of records within an approved tolerance is:

A record is “within tolerance” only if the organization has defined tolerances in advance. For example, a high-value serialized item may require exact agreement, while a low-value bulk material may permit a small quantity difference.

The formula is a proportion of records, not a measure of financial value. One missing low-value item and one missing high-value item each count as one inaccurate record.

6.2 Quantity and value accuracy

A quantity-based absolute variance rate can be expressed as:

This measure is meaningful when quantities are comparable and the denominator is positive. Absolute values prevent overages and shortages from canceling each other. For value control:

Here, unit cost must be consistently defined—for example, standard cost, moving average cost, or approved financial cost. The measure is not interchangeable with a net inventory gain or loss.

6.3 Shrinkage

Shrinkage is inventory loss not explained by authorized sales, production consumption, transfers, or other legitimate transactions. A basic period measure is:

Expected inventory should be based on a clearly defined opening balance plus authorized receipts minus authorized issues and other approved movements. The formula assumes the denominator is positive and that the period boundary and transaction cut-off are consistent. A negative result may indicate an overage, timing error, or unrecorded receipt rather than “negative shrinkage.”

Worked example

A warehouse begins a week with 2,400 units. It receives 900 units and issues 1,100 units. No approved transfers or adjustments occur. Expected closing inventory is:

2,400 + 900 − 1,100 = 2,200 units.

The physical count finds 2,178 units. The shortage is 22 units, so:

The result should not immediately be treated as theft. The organization should first test cut-off, unit conversions, unposted receipts, unrecorded damage, picking errors, and count quality.

6.4 Reconciliation

Reconciliation compares independent evidence and explains differences. A sound reconciliation process includes:

1. freeze or time-stamp the population being tested;
2. establish the book balance and source transactions;
3. perform a controlled physical count;
4. compare physical and recorded quantities;
5. classify discrepancies by cause;
6. obtain review and approval;
7. post authorized corrections;
8. retain evidence and monitor recurrence.

The objective is not merely to make the system equal the count. It is to identify why the difference occurred and prevent recurrence.

7. Cycle Counting and the Control of Discrepancies

Annual physical inventory counts provide a broad assertion at a point in time but may disrupt operations and detect problems late. Cycle counting counts selected inventory continuously throughout the year.

7.1 Designing a cycle-count program

A risk-based program considers value, criticality, demand, volatility, theft exposure, shelf life, prior error history, and regulatory importance. ABC classification is useful but insufficient. A low-value safety component can deserve more frequent counting than a high-value but stable item.

A practical policy may combine:

• frequent counts for high-value or high-risk items;
• scheduled counts for medium-risk items;
• periodic random counts for low-risk items;
• mandatory recounts after significant discrepancies;
• event-triggered counts after relocation, system migration, suspected theft, or process change.

Counting frequency should be based on risk and evidence, not an arbitrary desire to count everything equally.

7.2 Count methods

Blind counting hides the book quantity from the counter and reduces confirmation bias. Open counting displays the expected quantity and may be faster for routine verification, but it can cause counters to reproduce the record. A hybrid method may use blind first counts and open supervisory recounts.

Counters should record the physical quantity, unit, condition, location, lot or serial data, and any observations. They should not alter the book record during the count. Differences should be independently verified before adjustment.

7.3 Root-cause categories

Useful discrepancy categories include:

• receiving error;
• put-away or location error;
• picking or issue error;
• unrecorded return;
• unit-of-measure or master-data error;
• timing and cut-off error;
• damage, expiry, or scrap not processed;
• theft or unauthorized removal;
• counting error;
• system interface or integration failure;
• supplier or customer ownership error.

Pareto analysis of these categories helps management prioritize corrective action. An adjustment that hides a repeated receiving error is not an effective control response.

8. Internal Control and Audit-Ready Governance

Inventory controls should be designed using the principles of authorization, segregation, verification, safeguarding, and evidence.

8.1 Segregation of duties

Where practical, different people should be responsible for ordering, receiving, custody, recording, approving adjustments, and reviewing reports. Small organizations may not have enough personnel for complete segregation; compensating controls can include supervisory review, independent counts, access logs, and periodic management reconciliation.

8.2 Authorization and thresholds

Policies should define who may approve:

• emergency receipts and purchases;
• stock substitutions;
• write-offs and disposals;
• quantity and value adjustments;
• blocked or quarantined stock release;
• inventory transfers;
• cycle-count recounts.

Approval thresholds should reflect both value and risk. A low-value hazardous item may require greater control than a higher-value ordinary consumable.

8.3 Evidence and audit trail

Audit-ready evidence may include purchase orders, delivery notes, inspection records, photographs, barcode logs, count sheets, adjustment forms, disposal certificates, system histories, access records, and management review sign-offs. Evidence should be attributable, dated, complete, protected from unauthorized alteration, and retrievable.

A strong audit trail connects the physical event, source document, system transaction, authorization, and accounting impact. Electronic records should have role-based access, user identification, time stamps, change history, and appropriate retention controls.

8.4 Key performance and control indicators

Useful indicators include inventory accuracy, adjustment value, shrinkage rate, count completion, unresolved discrepancy age, receiving discrepancy rate, pick accuracy, location accuracy, quarantine dwell time, and percentage of transactions posted at the point of activity. Metrics should be disaggregated by facility, process, item class, shift, and cause; otherwise averages can conceal local failure.

9. Technology, Data, and Artificial Intelligence

Barcode scanning is relatively inexpensive and effective when labels, master data, and workflows are reliable. RFID can capture multiple tagged items without line-of-sight and may improve visibility in high-volume environments, although tag cost, interference, read accuracy, and process integration must be evaluated.

Warehouse-management systems can enforce locations, task sequences, status controls, and scanning. ERP integration is valuable only when interfaces are complete and timely. A sophisticated system cannot correct inaccurate item masters, poor training, or unauthorized physical movements.

Analytics can identify unusual adjustment patterns, repeated variances by location, negative balances, impossible transaction sequences, and users with abnormal activity. AI may support anomaly detection, image-assisted counting, demand-risk prioritization, or investigation triage. Its output should be treated as decision support rather than automatic proof of fraud. Models can reproduce biased data, generate false positives, or miss novel behavior. Human review, explainability, access control, and documented escalation remain necessary.

Technology should therefore be selected according to control objectives and total process economics, not novelty. A simple scan at the correct point is often more valuable than an advanced platform used after the event.

10. Illustrative Case: Meridian Components Distribution Centre

Background

Meridian Components supplies maintenance parts to industrial customers. Its ERP reported 98.2% inventory accuracy, yet customers experienced frequent stockouts. The warehouse also carried excess safety stock and recorded rising adjustment values.

Problem

An investigation found that the headline accuracy figure counted records within a tolerance of ±5 units. High-volume fasteners therefore appeared accurate even when quantities were materially wrong. In addition, receipts were posted at the end of shifts, returned parts were placed directly into available stock, and several storage locations contained mixed units of measure.

Analysis

A cross-functional team compared receiving logs, scan data, count sheets, adjustment records, and dispatch documents. It found four major causes:

1. receipt cut-off differences between the dock and ERP;
2. case-to-each conversion errors in the item master;
3. returns lacking inspection status;
4. picking from substitute locations without recording the movement.

The team recalculated accuracy using exact agreement for serialized parts and value-weighted variance for critical components. The reported performance declined, but the new measure was more decision-useful.

Approach and solution

Meridian introduced the following controls:

• scanning at receipt, put-away, pick, and dispatch;
• controlled conversion factors for cases, boxes, and individual units;
• quarantine locations for all returns pending inspection;
• prohibition of unrecorded substitute picking;
• blind cycle counts for high-risk items;
• daily review of negative balances and unposted receipts;
• adjustment approval based on both value and criticality;
• weekly root-cause review of the top discrepancy categories.

The company retained a tolerance-based operational measure but reported it alongside exact accuracy, absolute value variance, and unresolved discrepancy age.

Outcome

After three months, receiving cut-off errors and return-related discrepancies decreased substantially. Stockouts declined because planners could trust available balances, while excess purchasing was reduced after safety stocks were recalculated. The most important outcome was not the percentage improvement itself; it was the creation of a traceable process in which physical movements and system records were generated together.

Learning points

The case illustrates that:

• an accuracy percentage is meaningful only when its definition is explicit;
• tolerances should reflect item risk and value;
• inventory discrepancies are often process failures rather than isolated counting mistakes;
• returns and substitutions require status and authorization controls;
• technology improves control only when embedded in disciplined operations.

11. Challenges, Limitations, and Ethical Considerations

Perfect accuracy is costly and may be unnecessary. The appropriate target depends on item criticality, financial materiality, service requirements, safety, and regulatory exposure. Excessive counting can consume labor without addressing root causes; insufficient counting allows errors to accumulate.

Physical controls can also create ergonomic and safety risks. Heavy or hazardous materials require compliant handling, suitable storage, protective equipment, and trained personnel. Employees should never be pressured to bypass safety or inspection procedures to meet productivity targets.

Inventory adjustments affect performance reports, bonuses, taxes, financial statements, and sometimes customer commitments. Concealing shortages through unauthorized adjustments is unethical and may constitute fraud. Investigations should preserve evidence, respect due process, protect whistle-blowers, and avoid accusing individuals without corroboration.

Data privacy also matters. Scan systems, CCTV, and user analytics may monitor employee behavior. Organizations should communicate purpose, limit access, retain data appropriately, and comply with applicable employment and privacy law.

Best-Practice Framework

A mature inventory-control system generally follows these principles:

1. Define ownership, status, units of measure, and location before movement.
2. Record transactions at the point of physical activity.
3. Separate custody, recording, authorization, and review where feasible.
4. Use risk-based rather than purely uniform counting.
5. Investigate discrepancies before posting corrections.
6. Measure both record accuracy and value impact.
7. Treat returns, damage, quarantine, and disposal as controlled states.
8. Maintain complete, attributable, and retrievable evidence.
9. Review trends and root causes, not only monthly percentages.
10. Use technology to strengthen a sound process, not to compensate for an unsound one.

Chapter Summary

Inventory operations link physical custody with information integrity. Receiving must verify identity, quantity, condition, ownership, and status before stock becomes available. Storage requires controlled locations, clear identification, appropriate slotting, and segregation of usable, damaged, quarantined, expired, and customer-owned goods. Picking, issue, and dispatch should be authorized, traceable, and supported by item, location, lot, serial, and quantity verification. Returns must be inspected before re-entry into available stock.

Inventory accuracy should be defined carefully. Record accuracy measures the proportion of records within a stated tolerance, while absolute quantity or value variance measures the magnitude of error. Shrinkage calculations require consistent cut-off and a reliable expected balance. Cycle counting provides continuous assurance and should be risk-based, supported by blind counts or independent verification, and connected to root-cause analysis.

Audit-ready control depends on segregation of duties, authorization thresholds, safeguarding, system access controls, evidence, and documented reconciliation. Barcodes, RFID, warehouse-management systems, analytics, and AI can improve visibility and detect anomalies, but they cannot replace disciplined processes, competent people, ethical governance, and management accountability.

Key Terms

Inventory accuracy: The degree to which physical inventory agrees with recorded inventory within defined rules.

Cycle counting: Periodic counting of selected inventory rather than waiting for one annual count.

Shrinkage: Inventory loss not explained by authorized transactions or approved adjustments.

Reconciliation: Comparison of independent records or evidence to identify and explain differences.

Stock-keeping unit (SKU): A distinct item identity managed for inventory purposes.

Unit of measure: The standardized quantity basis used to receive, store, issue, and report an item.

Put-away: Movement of received goods to an assigned storage location.

Quarantine: Controlled status for goods awaiting inspection, decision, or disposition.

FIFO: First-in, first-out issue method based on receipt sequence.

FEFO: First-expiry, first-out issue method based on earliest expiry date.

Blind count: A physical count performed without showing the counter the book quantity.

Tolerance: An approved range within which physical and recorded quantities are treated as matching.

Adjustment: An authorized correction to an inventory record.

Cut-off: The rule determining the period in which a transaction is recorded.

Segregation of duties: Division of incompatible responsibilities among different people or controls.

Audit trail: Traceable evidence connecting an event, authorization, system entry, and outcome.

Master data: Controlled reference data such as SKU, unit, conversion, location, and status attributes.

Available inventory: Stock that is approved and eligible for allocation or issue.

Review Questions

Short Answer Questions

1. Why should quantity acceptance and quality acceptance be treated as separate receiving decisions?
2. Distinguish FIFO from FEFO.
3. What is the purpose of a blind inventory count?
4. Identify four common root causes of inventory discrepancies.
5. Why should an inventory adjustment include a reason code and authorization?

Long Answer Questions

1. Design a receiving process for a warehouse handling both ordinary materials and expiry-sensitive products.
2. Explain how location control, unit-of-measure discipline, and status segregation affect inventory accuracy.
3. Compare annual physical inventory counting with risk-based cycle counting.
4. Discuss the internal-control principles required for audit-ready inventory operations.
5. Evaluate the benefits and limitations of barcode, RFID, analytics, and AI in inventory control.

Analytical / Application Questions

1. A warehouse begins with 8,000 units, receives 3,500, issues 4,200, and records no approved adjustments. A physical count finds 7,250 units. Calculate expected closing inventory and shrinkage percentage, then identify at least three possible causes requiring investigation.
2. A company reports 99% inventory accuracy using a tolerance of ±10 units, but customers still experience stockouts of critical components. Critically evaluate the metric and propose a revised measurement framework.
3. Design a cycle-counting policy for a facility containing high-value serialized parts, low-value fasteners, hazardous chemicals, and customer-owned inventory. Explain frequency, counting method, evidence, and escalation rules.

Multiple Choice Questions with Answers

Question 1

Which control objective requires recorded inventory to physically exist?

A. Completeness

B. Existence

C. Cut-off

D. Authorization

Correct Answer: B. Existence

Question 2

Which issue method is normally most appropriate for expiry-sensitive products?

A. LIFO

B. Random issue

C. FEFO

D. Highest-cost-first

Correct Answer: C. FEFO

Question 3

A warehouse begins with 1,000 units, receives 300, and issues 450. What is the expected closing inventory?

A. 550 units

B. 700 units

C. 850 units

D. 1,750 units

Correct Answer: C. 850 units

Question 4

What is the principal purpose of a blind count?

A. To eliminate the need for authorization

B. To reduce confirmation bias from the recorded quantity

C. To permit immediate adjustment during counting

D. To replace all system records

Correct Answer: B. To reduce confirmation bias from the recorded quantity

Question 5

Which transaction should normally place returned goods into quarantine rather than immediately into available stock?

A. A return whose condition has not yet been inspected

B. A receipt that passed all required inspections

C. A completed dispatch with proof of delivery

D. A routine replenishment movement

Correct Answer: A. A return whose condition has not yet been inspected

Question 6

A record-level inventory accuracy measure counts a record as accurate when it is:

A. The highest-value item in the warehouse

B. Within the organization’s approved tolerance

C. Counted at least once per year

D. Located near the dispatch area

Correct Answer: B. Within the organization’s approved tolerance

Question 7

Which is an example of segregation of duties?

A. One employee orders, receives, records, and approves adjustments

B. All employees share one system password

C. Receiving, custody, recording, and adjustment approval are separated where feasible

D. Counters change records while performing the count

Correct Answer: C. Receiving, custody, recording, and adjustment approval are separated where feasible

Question 8

A warehouse has an expected balance of 2,200 units and a physical balance of 2,178 units. What is the shrinkage percentage?

A. 0.1%

B. 1.0%

C. 10.0%

D. 22.0%

Correct Answer: B. 1.0%

Question 9

Which is the strongest reason to investigate repeated inventory adjustments?

A. Adjustments always improve accuracy

B. Repeated adjustments may indicate an unresolved process or master-data failure

C. Adjustments remove the need for cycle counting

D. Adjustments prove supplier fraud

Correct Answer: B. Repeated adjustments may indicate an unresolved process or master-data failure

Question 10

Which statement about AI in inventory control is most accurate?

A. AI output is automatically proof of fraud

B. AI eliminates the need for physical counts

C. AI can support anomaly detection but requires human review and governance

D. AI makes item-master quality irrelevant

Correct Answer: C. AI can support anomaly detection but requires human review and governance