12. Digital Transportation Management and Artificial Intelligence

Introduction and Context

Transportation management has moved from a largely transactional activity—booking vehicles, preparing documents, and tracking deliveries—to an integrated, data-intensive managerial discipline. Modern transport decisions are made within networks involving shippers, carriers, brokers, ports, warehouses, customs agencies, customers, and technology providers. Each participant generates information about orders, capacity, location, status, cost, risk, and service performance. The managerial challenge is no longer simply to collect this information, but to connect, interpret, and use it quickly enough to improve decisions.

Digital transportation management refers to the use of interconnected software, data, communication networks, sensors, analytics, and automation to plan, execute, monitor, and improve transportation activities. Artificial intelligence (AI) extends these capabilities by identifying patterns, making predictions, recommending actions, and, within defined limits, executing decisions. The value of digitalisation does not arise from technology alone. It depends on data quality, process discipline, integration, governance, user adoption, and the ability of managers to balance efficiency with resilience, safety, compliance, and customer service.

This chapter examines the digital architecture of transportation and distribution management. It links transportation management systems (TMS) with enterprise resource planning (ERP), warehouse management systems (WMS), electronic data interchange (EDI), application programming interfaces (APIs), control towers, GPS, telematics, the Internet of Things (IoT), digital freight platforms, and blockchain. It then considers AI applications in forecasting, routing, estimated time of arrival (ETA) prediction, anomaly detection, and autonomous decision-making under human oversight.

Learning Objectives

After studying this chapter, students should be able to:

1. Explain the role and architecture of digital transportation management.

2. Evaluate TMS platforms and their integration with ERP and WMS environments.

3. Distinguish between EDI, APIs, control towers, telematics, IoT, digital freight platforms, and blockchain applications.

4. Apply analytical and AI methods to forecasting, routing, ETA prediction, and anomaly detection.

5. Assess the operational, financial, ethical, cybersecurity, and governance implications of automation.

6. Design a practical roadmap for implementing digital transportation capabilities with appropriate human oversight.

From Transaction Processing to Intelligent Transportation Management

A traditional transport process is often fragmented. A customer order may be recorded in an ERP system, inventory availability in a WMS, carrier capacity in a spreadsheet or carrier portal, shipment events in a telematics system, and freight invoices in a finance application. Manual re-entry creates delays and errors. More importantly, fragmented information prevents the organisation from seeing the shipment as one end-to-end process.

A digital transportation system creates a connected information flow. Demand and order data support planning; planning generates loads and routes; execution systems communicate with carriers and drivers; sensors report movement and conditions; the control tower presents exceptions; and settlement systems verify freight charges. The resulting feedback supports continuous improvement.

Digital maturity can be understood as a progression:

• Digitisation: paper records are converted into electronic records.
• Digital integration: systems exchange information automatically.
• Digital visibility: managers receive near-real-time views of shipments and network conditions.
• Digital optimisation: algorithms recommend improved plans using multiple constraints.
• Intelligent automation: systems predict events and execute approved decisions, while people govern exceptions and risk.

The progression is not necessarily linear. An organisation may have advanced vehicle tracking but weak master data, or a sophisticated TMS but poor carrier adoption. Consequently, technology evaluation must consider the complete operating model rather than isolated features.

The Digital Architecture of Transportation

Transportation Management Systems

A TMS is a software platform used to plan, procure, execute, monitor, audit, and analyse transportation. Core functions commonly include order management, shipment consolidation, carrier selection, rate management, tendering, route and load planning, appointment scheduling, track-and-trace, freight audit and payment, claims management, and performance reporting.

TMS platforms may be deployed on premises, through cloud software, or through a hybrid arrangement. Cloud systems generally support faster updates, scalability, and access across locations, although they create dependency on connectivity, vendor performance, and data security. A TMS should be evaluated against the organisation’s transport profile: modes used, shipment volume, geographic scope, complexity of rates, need for real-time visibility, carrier network, regulatory obligations, and integration requirements.

A useful evaluation framework considers five dimensions. First is functional fit: whether the system supports the actual planning and execution processes. Second is integration capability: whether it can exchange reliable data with ERP, WMS, carrier, telematics, customs, and financial systems. Third is analytical capability: whether it provides configurable metrics, scenario modelling, and usable dashboards. Fourth is scalability and usability: whether planners, carriers, drivers, and managers can use it effectively as volume grows. Fifth is governance and commercial risk: including cybersecurity, data ownership, service-level agreements, implementation effort, and total cost of ownership.

A TMS is not automatically an optimisation system. If order data, addresses, product dimensions, carrier rates, and transit times are inaccurate, the system may produce technically valid but operationally poor plans. Implementation therefore requires process standardisation, master-data governance, training, and change management.

ERP and WMS Integration

The ERP normally contains commercial, financial, procurement, order, and master-data information. The WMS manages inventory, storage, picking, packing, staging, and warehouse appointments. The TMS converts relevant order and inventory information into transport decisions and returns shipment status, cost, and delivery information to the enterprise.

Integration should be designed around business events rather than only around system screens. Typical events include order released, inventory allocated, shipment created, carrier tendered, tender accepted, vehicle departed, delivery attempted, proof of delivery received, and invoice approved. Each event should have a defined owner, data format, timing expectation, and exception procedure.

Poor integration produces familiar problems: duplicate shipments, vehicles arriving before goods are ready, incorrect delivery promises, unmatched freight invoices, and inconsistent customer status messages. Effective integration reduces rekeying, improves promise accuracy, and creates a common operational record.

EDI and APIs

Electronic data interchange is the structured computer-to-computer exchange of standard business documents. In transportation, EDI may carry load tenders, shipment status messages, advance shipment notices, invoices, and delivery confirmations. EDI is valuable where high-volume, repetitive, standardised transactions must be exchanged reliably among established trading partners.

APIs provide programmatic access between applications. They are often more flexible than traditional document exchange and can support near-real-time requests and responses, such as obtaining a freight quote, checking vehicle location, booking capacity, or requesting an updated ETA. APIs are particularly useful in platform ecosystems, although they require authentication, version control, monitoring, and clear responsibility for failures.

EDI and APIs are not mutually exclusive. A mature network may use EDI for contractual or high-volume transaction flows and APIs for dynamic visibility or pricing. The choice should reflect partner capability, transaction urgency, data complexity, integration cost, and governance requirements.

Control Towers

A transportation control tower is a coordinated visibility and decision-support capability that combines data from multiple internal and external sources. It is more than a dashboard. A genuine control tower applies business rules, analytics, workflow, and escalation to identify risks and coordinate responses across the network.

For example, a control tower may combine order priorities, inventory positions, weather alerts, port congestion, vehicle location, customer delivery windows, and carrier capacity. It can then highlight shipments likely to miss delivery commitments and propose alternatives. The human team decides whether to expedite, reroute, split an order, change the customer promise, or accept the risk.

Control towers can be centralised, regional, or virtual. Centralisation improves standardisation and network visibility, but local expertise remains important where regulations, customer requirements, or operating conditions differ. The principal danger is information overload. A control tower should prioritise exceptions according to financial impact, customer importance, safety, and recovery options.

Data from Vehicles, Assets, and the Physical Network

GPS provides geographic position derived from satellite-based positioning. It supports location tracking, route adherence, geofencing, and basic ETA calculation. Telematics combines location with vehicle and driver data such as speed, engine condition, fuel use, braking, idling, temperature, and driving hours. IoT extends this concept through connected sensors on vehicles, containers, pallets, packages, warehouses, and infrastructure.

The managerial value of these technologies depends on the decision they support. A temperature sensor is useful when a cold-chain process defines acceptable thresholds, escalation rules, corrective actions, and liability responsibilities. A location signal is useful when it changes customer communication, dock scheduling, or recovery planning. Collecting data without an operational response creates cost without proportional value.

Important design considerations include sensor accuracy, battery life, connectivity, calibration, data frequency, coverage, ownership, and interoperability. In international transport, a device may move through areas with different cellular networks and regulatory requirements. Data must also be associated with the correct shipment, asset, and time zone; otherwise, visibility becomes misleading.

Digital Freight Platforms and Blockchain

Digital freight platforms connect shippers and carriers through electronic marketplaces or collaborative networks. They may provide load posting, capacity discovery, digital tendering, instant or negotiated pricing, document management, payment, ratings, and tracking. Their advantages include broader market access, faster matching, reduced communication effort, and potentially improved asset utilisation.

However, platform participation can create concerns about price transparency, data ownership, carrier dependence, service quality, and the treatment of smaller transport providers. A platform should be evaluated not only by transaction speed but also by network liquidity, carrier qualification, dispute resolution, cybersecurity, and integration with existing systems.

Blockchain is a distributed ledger technology in which authorised participants share a tamper-evident record of transactions. In transportation, possible applications include chain-of-custody records, proof of provenance, automated contractual conditions, and trusted document exchange. Blockchain does not guarantee that an input is true. If a person or sensor records an incorrect event, the ledger may preserve the error. It can also introduce governance, privacy, scalability, and energy or infrastructure concerns. Blockchain is therefore most appropriate where multiple parties need a shared record and no single party is sufficiently trusted or accepted as the sole record keeper.

Artificial Intelligence in Transportation Decisions

AI in transportation includes machine learning, optimisation, natural language processing, computer vision, and rule-based automation. These techniques serve different purposes. Machine learning identifies relationships in historical data; optimisation searches for the best solution under constraints; natural language processing extracts meaning from documents or messages; computer vision interprets images; and rules enforce explicit operational policies.

Forecasting and Capacity Planning

Forecasting models estimate shipment volumes, lane demand, delivery time, cancellations, or capacity requirements. Inputs may include historical orders, seasonality, promotions, customer behaviour, weather, economic indicators, and operational calendars. Better forecasts help organisations position vehicles, negotiate capacity, schedule labour, and reserve warehouse or terminal appointments.

Forecast accuracy should be assessed using appropriate measures such as mean absolute error, mean absolute percentage error where suitable, or forecast bias. A forecast can have a low average error but systematically underestimate peak demand. Managers should therefore examine errors by lane, customer, time period, and demand segment. Forecasts are decision inputs, not promises; planners must retain the ability to incorporate known events that are absent from historical data.

Routing and Load Optimisation

Vehicle routing problems involve assigning shipments to vehicles and sequences of stops while considering capacity, time windows, driver regulations, costs, service priorities, road restrictions, and uncertainty. Exact optimisation can become computationally difficult as the number of stops and constraints increases. Practical systems use mathematical programming, heuristics, metaheuristics, constraint programming, or hybrid approaches.

The best route is not necessarily the shortest route. A slightly longer route may reduce missed delivery windows, balance driver workload, avoid tolls, protect high-priority customers, or improve resilience. A responsible optimisation model makes objectives and constraints visible. If planners cannot understand why a recommendation was made, adoption and accountability may suffer.

ETA Prediction

ETA prediction combines planned route information with real-time location, historical travel times, traffic, weather, stops, border conditions, and driver or vehicle behaviour. A useful ETA should include confidence or uncertainty information where possible. The system should distinguish between a normal variation and a material risk to the promised delivery window.

ETA models must be recalibrated as network conditions change. A model trained on normal operations may perform badly during exceptional congestion, strikes, disasters, or new regulatory restrictions. Human planners should be able to override an ETA when they possess credible information not represented in the data.

Anomaly Detection

Anomaly detection identifies unusual patterns such as unexplained route deviation, excessive dwell time, abnormal fuel use, temperature excursions, repeated failed delivery attempts, suspicious access to cargo, or freight charges outside expected ranges. Methods include statistical thresholds, clustering, classification models, and sequential pattern analysis.

Anomaly detection should not treat every unusual event as misconduct. A route deviation may result from a road closure; a long dwell time may reflect a customer’s appointment process. Therefore, alerts require context, prioritisation, investigation workflows, and feedback so that the system learns which alerts are meaningful.

Autonomous Decisions with Human Oversight

Automation can range from assistance to full execution. A system may display a recommendation, automatically tender a load within approved rules, or autonomously reroute a shipment. The appropriate level depends on risk, reversibility, regulatory requirements, financial exposure, and the quality of available data.

A human-in-the-loop model requires clear thresholds. Low-risk decisions, such as selecting among pre-approved carriers for routine shipments, may be automated. High-risk decisions involving dangerous goods, medical products, major service failures, discrimination risks, or significant financial commitments should require human approval. Every automated action should be traceable through logs showing the data used, rule or model applied, decision taken, and responsible authority.

Implementation as an Operating-Model Change

Digital transformation should begin with business problems rather than product demonstrations. A useful implementation sequence is:

1. Establish objectives and baseline performance for cost, service, safety, emissions, and working capital.

2. Map current processes, data flows, decision rights, and pain points.

3. Assess data quality, integration readiness, cybersecurity, and partner capability.

4. Select a target architecture and prioritise high-value use cases.

5. Pilot within a controlled lane, business unit, or carrier group.

6. Validate benefits, user experience, model performance, and exception handling.

7. Scale through governance, training, supplier engagement, and continuous improvement.

Key performance indicators may include transport cost per shipment or unit, tender acceptance, planned-versus-actual transit time, on-time in-full delivery, empty kilometres, vehicle utilisation, dwell time, ETA accuracy, claims, invoice accuracy, alert resolution time, and emissions per shipment. Metrics should be balanced. Driving down cost while increasing damage, unsafe driving, or failed deliveries is not genuine improvement.

Illustrative Case: A Regional Retailer’s Digital Control Tower

Background

A fictional regional retailer operates distribution centres and stores across several countries. It uses an ERP for orders, a WMS for distribution centres, separate carrier portals, and spreadsheets for daily transport planning. The retailer has regular visibility gaps and relies on telephone calls when deliveries are delayed.

Situation and Problem

The retailer experiences missed delivery windows, high planner workload, inconsistent carrier performance, and disputed freight invoices. Management considers purchasing an AI-enabled TMS immediately, but an initial diagnostic reveals inconsistent store codes, incomplete delivery time windows, outdated carrier rates, and limited tracking coverage.

Analysis

The primary problem is not the absence of AI. It is fragmented data and unclear processes. The project team maps order release, warehouse readiness, carrier tendering, tracking, proof of delivery, and invoice approval. It identifies that many “late deliveries” are actually caused by late warehouse staging or incorrect appointment information. This distinction is essential because a routing model cannot solve a warehouse-release problem.

Approach and Solution

The retailer first standardises location and carrier master data, defines shipment-status events, and establishes API and EDI requirements. It integrates ERP orders and financial data with the WMS and a cloud TMS. Selected carriers provide telematics feeds, while smaller carriers use a mobile application or structured status messages.

A control tower is introduced for high-priority store deliveries. It uses rules and predictive ETA analytics to identify shipments likely to miss their windows. Routine carrier selection within approved rates is automated; exceptions involving premium freight, temperature-sensitive goods, or major customer impact require planner approval. The system records the reasons for overrides and uses them to improve planning rules.

Outcome

In this illustrative case, the retailer improves shipment visibility, reduces manual status calls, and separates warehouse-related delays from transport-related delays. Planners spend less time searching for information and more time resolving important exceptions. The organisation does not claim that the algorithm eliminates disruption; rather, it improves the speed and quality of response.

Learning Points

The case demonstrates that digital value depends on data governance and process clarity. A control tower must support decisions, not merely display events. Automation should begin with bounded, repeatable decisions, while high-impact exceptions remain subject to human review. Finally, benefits must be measured against a baseline and assessed across cost, service, resilience, safety, and employee workload.

Risks, Governance, and Professional Responsibility

Digital transportation systems expand the attack surface of the organisation. Threats include unauthorised access, ransomware, manipulation of location data, compromised vehicle devices, fraudulent carrier identities, and leakage of commercially sensitive rates. Controls should include identity management, encryption, network segmentation, secure API gateways, patching, vendor assessment, backups, incident response, and audit logging.

Data governance covers accuracy, ownership, access, retention, lineage, and permitted use. Personal data may arise from driver locations, behaviour, communications, and working hours. Organisations must comply with applicable privacy, employment, transport, customs, and safety requirements. Driver monitoring should have a legitimate purpose, proportionality, transparency, and appropriate access restrictions.

AI introduces additional governance questions. Models can reproduce historical bias, fail under changed conditions, or create false confidence. Procurement teams should require documentation of training data, performance by relevant segment, explainability appropriate to the decision, drift monitoring, fallback procedures, and responsibility for errors. Human oversight must be substantive rather than symbolic: staff need authority, time, information, and training to challenge automated recommendations.

Common implementation mistakes include buying software before defining processes, treating visibility as the same as control, ignoring smaller carriers, automating poor master data, measuring only cost, accepting excessive alerts, and failing to plan for system outages. A resilient design includes manual fallbacks, offline procedures, data reconciliation, and tested recovery arrangements.

Emerging Directions

Transportation management is moving toward composable digital architectures in which specialised services connect through APIs rather than one monolithic application. Digital twins may support simulation of network changes, facility closures, capacity shortages, and disruption scenarios. Generative AI may assist with transport-document interpretation, natural-language querying, exception summaries, and planner communication, but outputs require validation because fluent language does not guarantee factual accuracy.

Connected and increasingly automated vehicles, drones, robotics, and advanced driver-assistance systems may change the economics and risk profile of distribution. Their adoption will be shaped by infrastructure, regulation, labour availability, insurance, cybersecurity, and public acceptance. Sustainability analytics is also becoming more integrated with routing and procurement, enabling decisions that consider emissions alongside cost and service.

The strategic direction is therefore not “replace managers with algorithms.” It is to build a transportation organisation in which reliable data, capable people, and well-governed automation reinforce one another. Competitive advantage will depend on the quality of decisions and the resilience of the digital operating model.

Best-Practice Principles

Effective digital transportation management generally follows these principles:

• Start with measurable business outcomes and a clear baseline.
• Treat master data and event definitions as strategic assets.
• Design integration around end-to-end processes and business events.
• Select technology according to network complexity and user needs.
• Use exception-based management rather than indiscriminate monitoring.
• Pilot bounded AI use cases before expanding autonomous authority.
• Maintain human accountability for high-impact decisions.
• Evaluate performance across cost, service, safety, resilience, sustainability, and fairness.
• Build cybersecurity, privacy, auditability, and recovery into the architecture.
• Involve carriers, drivers, warehouse teams, customers, and planners in design.

Chapter Summary

Digital transportation management connects planning, execution, visibility, settlement, and improvement through integrated technologies. TMS platforms coordinate transport decisions, while ERP and WMS integration links commercial orders, inventory readiness, and financial control. EDI supports structured document exchange; APIs enable flexible, near-real-time connectivity; control towers combine visibility with prioritised intervention. GPS, telematics, and IoT provide operational data, while digital freight platforms improve market coordination and blockchain may support shared records in carefully selected situations.

AI adds forecasting, routing optimisation, ETA prediction, anomaly detection, and controlled autonomous decisions. Its effectiveness depends on data quality, appropriate models, transparent objectives, and human oversight. Successful implementation is an organisational transformation involving governance, cybersecurity, process redesign, partner participation, training, and balanced performance measurement. The central managerial task is to use digital capabilities to improve decisions without surrendering accountability, resilience, safety, or professional judgement.

Key Terms

Artificial Intelligence: Computational methods that perform tasks involving prediction, classification, optimisation, language, perception, or decision support.

API: An application programming interface that enables software systems to exchange data or invoke functions.

Blockchain: A distributed, tamper-evident ledger shared among authorised participants.

Control Tower: An integrated visibility and decision-support capability that prioritises exceptions and coordinates responses.

Digital Freight Platform: An online network connecting shippers and carriers for capacity, transactions, documents, and visibility.

EDI: Electronic data interchange using structured formats to exchange standard business documents between systems.

ETA: Estimated time of arrival, usually calculated from planned and real-time transport information.

ERP: Enterprise resource planning software managing integrated commercial, financial, procurement, and operational records.

Geofencing: Defining a virtual geographic boundary that triggers an event when an asset enters or leaves an area.

IoT: A network of connected physical objects that collect and exchange data through sensors and communication systems.

Machine Learning: Algorithms that identify patterns in data and use them to make predictions or classifications.

Master Data: Stable reference information such as locations, products, carriers, rates, and vehicle attributes.

Telematics: Integrated collection and communication of vehicle, driver, location, and operating-condition data.

TMS: Transportation management system used to plan, execute, monitor, audit, and analyse transport.

WMS: Warehouse management system controlling inventory, storage, picking, packing, and warehouse execution.

Anomaly Detection: Identification of events or patterns that differ materially from expected behaviour.

Human-in-the-Loop: A governance arrangement in which people review, approve, override, or supervise automated decisions.

Digital Twin: A data-based representation of a physical network or process used for monitoring, simulation, or optimisation.

Review Questions

Short Answer Questions

1. Distinguish digitisation, digital integration, and intelligent automation in transportation.

2. State four core functions of a transportation management system.

3. How do EDI and APIs differ in transportation information exchange?

4. What types of information can telematics provide beyond vehicle location?

5. Why is human oversight necessary in high-impact AI transportation decisions?

Long Answer Questions

1. Explain how ERP, WMS, and TMS integration supports end-to-end transportation management.

2. Evaluate the role of a control tower and distinguish it from a simple shipment dashboard.

3. Discuss the applications and limitations of AI in routing, ETA prediction, and anomaly detection.

4. Propose an implementation roadmap for a company beginning digital transportation transformation.

5. Examine the cybersecurity, privacy, data-quality, and accountability issues associated with connected transportation systems.

Analytical / Application Questions

1. A carrier’s tracking feed shows repeated route deviations, but the region has severe weather and road closures. Design an anomaly-management process that avoids both false alarms and missed risks.

2. A retailer wants to automate carrier selection and premium-freight approval simultaneously. Which decision should be automated first, under what conditions, and why?

3. A company has an advanced TMS but inconsistent location codes, incomplete delivery windows, and poor carrier adoption. Analyse why expected AI benefits are unlikely and recommend priorities for the first year.

22. Multiple Choice Questions with Answers

Question 1

Which statement best describes the primary role of a transportation management system?

A. It controls only warehouse picking and storage activities.

B. It plans, executes, monitors, audits, and analyses transportation activities.

C. It replaces all carrier contracts with blockchain records.

D. It provides satellite signals without managing shipments.

Correct Answer: B. It plans, executes, monitors, audits, and analyses transportation activities.

Question 2

A company wants near-real-time access to carrier capacity and updated freight quotes. Which integration method is generally most suitable?

A. Manual spreadsheet exchange

B. Paper delivery documentation

C. API-based integration

D. Annual financial reporting

Correct Answer: C. API-based integration

Question 3

What makes a transportation control tower more than a basic tracking dashboard?

A. It displays every available data point without prioritisation.

B. It combines visibility with rules, analytics, workflows, and exception management.

C. It eliminates the need for carrier communication.

D. It records only completed deliveries.

Correct Answer: B. It combines visibility with rules, analytics, workflows, and exception management.

Question 4

Why might an AI routing model select a route that is not the shortest?

A. The model cannot consider constraints.

B. The longer route may better satisfy time windows, safety, cost, or service priorities.

C. Short routes are never permitted in transportation.

D. Routing models use only vehicle colour as an input.

Correct Answer: B. The longer route may better satisfy time windows, safety, cost, or service priorities.

Question 5

Which example is most clearly an anomaly-detection application?

A. Printing a standard bill of lading

B. Identifying unexplained temperature excursions in a cold-chain shipment

C. Entering a known customer address into an ERP system

D. Issuing a routine purchase order

Correct Answer: B. Identifying unexplained temperature excursions in a cold-chain shipment

Question 6

What is a major limitation of blockchain in transportation?

A. It cannot preserve transaction records.

B. It guarantees that every recorded input is factually correct.

C. It may preserve incorrect data if the original input is wrong.

D. It is useful only for vehicle navigation.

Correct Answer: C. It may preserve incorrect data if the original input is wrong.

Question 7

Which implementation priority is most appropriate before deploying advanced AI in a fragmented transport network?

A. Standardise master data and define reliable business events.

B. Automate all premium-freight approvals immediately.

C. Remove all human planners from the process.

D. Collect maximum data without defining its use.

Correct Answer: A. Standardise master data and define reliable business events.

Question 8

A model predicts that a shipment will miss its delivery window, but a planner knows that a recently opened alternative route is not in the model’s data. What is the best governance response?

A. Ignore the planner because models cannot be challenged.

B. Allow an informed override, record the reason, and use the feedback for model improvement.

C. Delete the shipment from the system.

D. Automatically cancel the delivery.

Correct Answer: B. Allow an informed override, record the reason, and use the feedback for model improvement.

Question 9

Which performance approach best evaluates digital transportation transformation?

A. Measure transport cost only.

B. Measure software purchase price only.

C. Balance cost, service, safety, resilience, sustainability, and workload outcomes.

D. Measure the number of dashboard screens created.

Correct Answer: C. Balance cost, service, safety, resilience, sustainability, and workload outcomes.

Question 10

Why should high-impact automated decisions retain human approval?

A. Algorithms are always unable to process data.

B. Human review supports accountability where risks, uncertainty, reversibility, or regulatory consequences are significant.

C. Human approval guarantees that no disruption will occur.

D. Automation has no value in routine decisions.

Correct Answer: B. Human review supports accountability where risks, uncertainty, reversibility, or regulatory consequences are significant.