Digital, Platform, Data and AI-Enabled Models

Introduction

Digital technologies have changed how ventures create, deliver and capture value. A software firm can serve millions of customers without owning conventional distribution infrastructure; a platform can coordinate transactions among users who do not belong to the same organization; cloud services can convert fixed technology investments into variable operating costs; and data and artificial intelligence (AI) can improve decisions, automate work and personalize experiences.

These possibilities do not automatically produce viable businesses. Digital ventures still require a coherent value proposition, identifiable customers, reliable operations, sustainable revenues and defensible capabilities. In addition, they must manage cybersecurity, privacy, algorithmic bias, regulatory obligations and the social consequences of automation. The central managerial task is therefore not simply to “add technology,” but to design an integrated business model in which technology strengthens customer value and economic viability.

This chapter examines digital, omnichannel, platform, ecosystem, cloud, data and AI-enabled models. It focuses on how entrepreneurs can design and evaluate these models while addressing network effects, governance, responsible innovation and operational risk.

Learning Objectives

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

1. Distinguish digital products, digital business models, platforms and ecosystems.
2. Design an omnichannel customer and operating model.
3. Explain direct and indirect network effects and their strategic implications.
4. Evaluate cloud, automation, data monetization and AI-enabled value creation.
5. Select meaningful performance metrics for digital and platform ventures.
6. Identify cybersecurity, privacy, governance and responsible-AI requirements.
7. Analyze a digital business model and recommend improvements.

1. Digital Business Models and Digital Value Creation

A digital business model uses digital technologies as a significant part of its value proposition, value-creation process, customer interface or revenue logic. Digitalization may affect an existing model incrementally, or it may enable an entirely new configuration of activities.

Examples include:

• A traditional retailer adding online ordering and home delivery.
• A software company selling access through a subscription rather than a perpetual license.
• A marketplace matching independent sellers with buyers.
• A data service converting operational information into forecasts for business customers.
• An AI-enabled application automating a previously human-intensive task.

Digital value creation commonly arises through six mechanisms:

1. Convenience and accessibility: Customers can transact at any time and from many locations.
2. Reduced search and transaction costs: Digital interfaces make comparison, matching and payment faster.
3. Personalization: Data allows products, recommendations and communications to be adapted to user needs.
4. Scalability: Software and digital content can often be replicated at low marginal cost.
5. Coordination: Platforms coordinate distributed users, suppliers, workers or assets.
6. Learning: Usage data can improve products, forecasts and operational decisions.

Digital models nevertheless have costs. Customer acquisition can be expensive, data infrastructure requires investment, service reliability is essential, and competitors can imitate features rapidly. A digital model should therefore be assessed through the same fundamental questions as any other business model:

• Who is the target customer?
• What problem is solved?
• What resources and activities create the solution?
• How does the venture reach customers?
• Who pays, how much and when?
• Which costs increase with growth?
• What risks could prevent sustainable value capture?

2. Omnichannel Models

A channel is a route through which a firm communicates with customers, sells products, delivers service or receives feedback. A multichannel firm uses several channels, such as a physical store, website, mobile application and social-media account. An omnichannel model goes further: it integrates channels so that customers experience the firm as one connected system.

For example, a customer may discover a product through social media, inspect it through a mobile application, purchase it online, collect it from a store, and obtain support through a call centre. An omnichannel model attempts to preserve consistent identity, product information, pricing logic, customer records and service quality across these interactions.

Important design decisions include:

• Customer identity: Can the firm recognize a customer across channels with appropriate consent?
• Inventory visibility: Are stock records updated sufficiently to support accurate promises?
• Order orchestration: Can orders be fulfilled from stores, warehouses or partners efficiently?
• Channel conflict: How will the firm manage competition between direct sales and intermediaries?
• Experience consistency: Are prices, policies, accessibility and service standards coherent?
• Data governance: Are customer data collected lawfully and used only for legitimate purposes?

Omnichannel integration is not always economically desirable. A small venture may gain more from one reliable digital channel than from several poorly integrated channels. Channel breadth should follow customer needs and operational capability, not technology fashion.

3. Platforms and Ecosystems

A pipeline business creates value through a relatively linear sequence: inputs are transformed into products or services and delivered to customers. A platform business creates value by enabling interactions between two or more participant groups. Examples include buyers and sellers, riders and drivers, hosts and guests, advertisers and audiences, or developers and users.

The platform usually provides rules, search, identity, matching, payment, reputation mechanisms and dispute resolution. It may not own the assets used in every transaction. An ecosystem is broader than a platform: it includes the platform, participants, complementary products, infrastructure, standards, regulators and other organizations whose activities jointly shape value creation.

3.1 Platform Types

Platforms must solve a chicken-and-egg problem: one participant group may not join until the other is present. A marketplace needs sellers to attract buyers, but sellers need buyers before joining is worthwhile. Entrepreneurs may address this problem through subsidies, direct recruitment, exclusive supply, geographic concentration or a narrow initial niche.

3.2 Network Effects

A network effect exists when the value received by a user changes as participation changes. A direct network effect occurs when users benefit from more users of the same type—for example, communication services. An indirect or cross-side network effect occurs when growth on one side increases value for another side—for example, more buyers attract sellers, while more sellers improve buyer choice.

Network effects are not always positive. Congestion, spam, fraud, low-quality content, privacy concerns and excessive complexity can reduce user value. Platform growth therefore requires both participant acquisition and quality governance.

A simplified platform adoption model can express user value as:

Here, V_i is the net value perceived by participant i, B_i(n) is the benefit generated by the relevant participant population n, P_i is the price or fee, and C_i represents participation costs such as time, learning, risk and privacy costs. The expression is conceptual rather than a universal demand equation. It is useful when comparing platform designs, provided that benefits and costs are defined consistently.

A platform becomes more attractive when network-generated benefits exceed price and participation costs. However, a large user count is not sufficient. Engagement, successful matches, trust and retention are often more meaningful than registrations.

3.3 Platform Governance

Platforms govern interactions through:

• Entry rules and participant verification.
• Search-ranking and recommendation systems.
• Pricing and commission policies.
• Ratings, reviews and reputation systems.
• Content, safety and conduct standards.
• Data-access and interoperability rules.
• Appeals, refunds and dispute-resolution mechanisms.

Governance involves a trade-off. Strict controls can improve safety and quality but may discourage participation or innovation. Weak controls can accelerate early growth but create fraud, discrimination, unsafe services or reputational damage. Governance rules should be transparent, proportionate, enforceable and subject to review.

4. Cloud Services and Automation

Cloud computing provides on-demand access to computing, storage, databases, networking and software through shared technical infrastructure. Its commercial significance lies in the ability to convert some technology expenditure from large upfront commitments into scalable operating expenditure.

Common service categories are:

• Infrastructure as a Service (IaaS): Virtual machines, networks and storage.
• Platform as a Service (PaaS): Development and deployment environments.
• Software as a Service (SaaS): Complete applications delivered to users.
• Function or serverless services: Code executed in response to events without the customer managing servers directly.

Cloud adoption can improve speed, flexibility and geographic reach, but it introduces dependence on providers, configuration risks, data-transfer costs, service outages and possible regulatory constraints on data location. A responsible cloud strategy includes access controls, encryption, backup, resilience testing, supplier due diligence and an exit or portability plan.

Automation uses rules, software or machines to perform tasks with limited human intervention. Robotic process automation may transfer information between systems; workflow automation may route approvals; machine learning may classify cases or forecast demand. Automation should begin with a clearly understood process. Automating a poorly designed process can make errors faster and harder to detect.

A useful implementation sequence is:

1. Map the current process and identify its purpose.
2. Measure time, cost, error rates and customer impact.
3. Remove unnecessary steps before automating.
4. Define exception handling and human escalation.
5. Test accuracy, security and accessibility.
6. Monitor outcomes after deployment.

5. Data as a Strategic Resource

Data may improve operations, customer experience and strategic decision-making. Potential data-enabled offerings include benchmarking, predictive maintenance, demand forecasts, fraud detection, market intelligence and personalized recommendations. Data can also support a feedback loop: usage generates data, analysis improves the service, and an improved service attracts or retains users.

Data is not automatically an asset of unlimited value. Its usefulness depends on quality, timeliness, completeness, lawful access, interoperability, analytical capability and the decision it supports. Poorly defined or biased data can produce confident but harmful decisions.

5.1 Data Monetization Models

1. Direct sale: A firm sells datasets or reports, subject to legal and contractual restrictions.
2. Insight sale: The firm sells analysis rather than raw data.
3. Data-enhanced product: Data improves an existing product or service.
4. Usage-based service: Customers pay for access to data processing or predictions.
5. Advertising or recommendation model: Data improves matching between audiences and offers.
6. Internal productivity model: Data creates value through reduced waste, better planning or lower risk rather than external revenue.

A strong data proposition explains whose data are used, what consent or legal basis applies, how the data are protected, what decision or benefit results, and why customers cannot easily obtain the same value elsewhere.

5.2 Data-Driven Performance Metrics

Digital ventures should distinguish activity metrics from outcome metrics. Downloads, registered accounts and website visits measure activity; successful transactions, retention, contribution margin and customer outcomes measure performance.

Customer acquisition cost may be expressed as:

where M is attributable marketing and sales expenditure during a defined period, and N is the number of new paying customers acquired during that same period. The calculation assumes that expenditure and acquisitions are measured over comparable periods and that attribution rules are stated.

A simple customer lifetime value estimate for a subscription business is:

where ARPU is average revenue per user per period, GM is the gross-margin ratio, and c is the periodic customer-churn rate. This steady-state approximation assumes relatively stable revenue, margin and churn, and ignores discounting, expansion revenue, acquisition costs and changing behaviour. It should not be treated as a precise valuation.

For example, if monthly average revenue is $20, the gross-margin ratio is 0.75 and monthly churn is 0.05, then:

LTV = (20 × 0.75) ÷ 0.05 = $300

If the venture’s customer acquisition cost is $120, the simplified LTV-to-CAC ratio is 2.5. Interpretation requires caution because the estimate excludes overhead, capital costs and uncertainty.

6. Artificial Intelligence in Business Models

AI-enabled models use computational systems to perform tasks commonly associated with human intelligence, including prediction, classification, language processing, recommendation, perception and content generation. AI can create value in three broad ways:

• Decision augmentation: Providing information or recommendations to employees and customers.
• Process automation: Performing repeatable tasks or routing cases.
• Product transformation: Making intelligence itself part of the customer offering.

Examples include demand forecasting, fraud detection, dynamic recommendations, customer-service assistants, document extraction, quality inspection and generative design tools.

AI value creation depends on a complete system, not merely a model. The system includes data collection, labelling, model development, interfaces, human workflows, monitoring, security and governance. A technically accurate model can still fail if users do not trust it, if data are unavailable at decision time, or if the process does not assign responsibility for errors.

6.1 AI Design Questions

Entrepreneurs should ask:

1. What decision or task is being improved?
2. Is AI superior to a rule, workflow or human process for this task?
3. What data are required, and are they lawful and representative?
4. What is the cost of false positives and false negatives?
5. Does the system require explanation, human review or appeal?
6. How will performance drift be detected?
7. Who is accountable when the system causes harm?

For a binary classification system, common evaluation measures include precision and recall. If TP denotes true positives, FP false positives and FN false negatives:

Precision answers, “Of the cases classified as positive, how many were actually positive?” Recall answers, “Of the actual positive cases, how many were identified?” The appropriate balance depends on context. In fraud screening, excessive false alarms may frustrate customers; in safety screening, missing a dangerous case may be more serious. These formulas assume clearly defined classes and reliable ground-truth labels.

Generative AI introduces additional concerns, including fabricated outputs, confidential-data leakage, copyright disputes, prompt manipulation and inconsistent responses. Human review, restricted data handling, source verification and use-case boundaries are essential.

7. Cybersecurity, Privacy and Responsible AI

Cybersecurity protects the confidentiality, integrity and availability of information and systems. A digital venture should identify assets, threats, vulnerabilities, impacts and controls. Common threats include phishing, credential theft, ransomware, insecure application programming interfaces, supply-chain compromise, insider misuse and denial-of-service attacks.

Basic controls include:

• Multi-factor authentication.
• Least-privilege access.
• Encryption in transit and at rest.
• Secure software development and code review.
• Patch and vulnerability management.
• Logging, monitoring and anomaly detection.
• Tested backups and incident-response plans.
• Employee training and supplier risk assessment.

Privacy is concerned with appropriate collection, use, disclosure and retention of personal information. Good practice includes purpose limitation, data minimization, accuracy, transparency, access and correction rights, retention limits and secure deletion. Legal requirements vary by jurisdiction, so entrepreneurs must obtain specialist advice rather than assume that one global approach is sufficient.

Responsible AI extends these principles to automated and semi-automated decisions. Core principles include:

• Validity and reliability: The system performs its intended task under relevant conditions.
• Fairness: Outcomes do not produce unjustified discrimination.
• Transparency: People receive understandable information about relevant system use.
• Explainability: Decisions can be interpreted sufficiently for the context.
• Human oversight: Responsible people can intervene, review and appeal.
• Privacy and security: Data and models are protected.
• Accountability: Roles, records and remedies are clearly assigned.
• Sustainability: Energy, infrastructure and social costs are considered.

Responsible AI is not a one-time checklist. It is a lifecycle practice involving impact assessment before deployment, testing during development, monitoring in operation and review when the use case or environment changes.

8. Designing an Integrated Digital Model

A practical design process can integrate the Business Model Canvas with technology and governance analysis.

Step 1: Define the user problem

Specify the customer, context, unmet need and evidence of willingness to adopt or pay. Avoid beginning with a technology looking for a problem.

Step 2: Identify participants and interactions

For a platform, distinguish each side, the interaction being enabled and the value exchanged. Determine whether the venture is a marketplace, infrastructure provider, community, innovation platform or combination.

Step 3: Map the value architecture

Show how data, software, people, partners, cloud resources and physical assets combine to deliver the promise. Identify bottlenecks and dependencies.

Step 4: Select revenue and pricing logic

Possible approaches include subscriptions, transaction fees, advertising, licensing, usage-based pricing, freemium conversion and outcome-based pricing. Pricing should reflect willingness to pay, marginal cost, cross-side network effects and strategic objectives.

Step 5: Design trust and governance

Specify identity verification, safety standards, moderation, dispute handling, privacy notices, consent, access control and accountability.

Step 6: Build a minimum viable system

A minimum viable product should test the riskiest business assumptions, not merely provide the smallest set of features. For a platform, the critical test may be successful matching and repeat participation rather than app downloads.

Step 7: Measure and learn

Track acquisition, activation, engagement, retention, successful interactions, contribution margin, service quality, security events and fairness indicators. Use experiments carefully: short-term engagement gains may damage trust or long-term retention.

Illustrative Case: GreenBasket Local Commerce Platform

Background

GreenBasket is an illustrative early-stage venture connecting independent food producers with urban households. Producers have excess seasonal inventory, while households want convenient access to local products. Existing arrangements are fragmented: producers rely on weekend markets, and customers use multiple messaging channels to place orders.

Problem

The venture faces three connected problems. Producers need demand visibility and reliable payment; customers need convenient discovery and delivery; and the venture must achieve enough order density to make logistics economical. A basic online catalogue alone would not solve coordination, trust or fulfilment.

Analysis

GreenBasket operates a two-sided platform. Producers benefit from customer access and demand data; households benefit from variety, comparison and coordinated delivery. The platform faces a local chicken-and-egg problem. If only a few producers join, customers see little choice. If only a few customers join, producers receive insufficient orders.

The initial strategy is to focus on one neighbourhood, recruit a credible anchor producer, and use scheduled weekly delivery windows. This concentrates demand geographically and reduces logistics complexity. Producer verification, product standards, refund rules and transparent ratings support trust.

Approach and Solution

GreenBasket adopts the following model:

• Customers browse and order through a web application.
• Producers upload availability through a simple supplier portal.
• The system aggregates orders before a weekly cut-off.
• A cloud-based inventory database tracks supply and substitutions.
• Demand forecasts help producers plan harvest and production.
• GreenBasket charges producers a transaction commission and customers a delivery fee.
• Customer data are used for fulfilment and opted-in recommendations, not sold as raw personal data.
• Automated fraud checks flag unusual payment activity, while staff review disputed cases.

The venture deliberately avoids fully autonomous pricing at launch. Dynamic pricing could reduce waste, but it might also confuse customers or disadvantage smaller producers. A transparent discount policy is used while demand data are accumulated.

Outcome

In the first pilot, management evaluates successful repeat orders, on-time delivery, spoilage reduction, producer retention, contribution margin per route and customer complaints. Registrations are treated as a secondary measure. The pilot reveals that demand density, rather than catalogue size, is the main driver of economic viability. It also shows that producer training is necessary because inconsistent stock updates create customer disappointment.

Learning Points

1. A platform’s value depends on successful interactions, not participant counts alone.
2. Geographic concentration can help overcome early coordination and logistics problems.
3. Data quality is an operational responsibility shared with participants.
4. Governance and trust mechanisms are part of the product.
5. Automation should be introduced according to risk and process maturity.
6. Responsible data use can be a source of differentiation rather than only a compliance cost.

Challenges and Limitations

Digital and AI-enabled models face several structural limitations:

• Scalability may be deceptive: Revenue can grow faster than capability, causing service failures.
• Winner-take-most dynamics are uncertain: Network effects can be weak, reversible or constrained by multi-homing, regulation and user distrust.
• Switching costs can harm users: Lock-in may reduce choice and encourage exploitative pricing.
• Data advantages may decay: Data become obsolete, competitors can obtain alternatives, and privacy rules may restrict use.
• Automation can displace work: Productivity gains may create transition costs and require reskilling.
• Cyber incidents can destroy trust: A single breach may impose legal, financial and reputational costs.
• AI can reproduce social inequalities: Historical data may encode discriminatory practices.
• Cloud dependency creates concentration risk: Dependence on a small number of providers can affect resilience and bargaining power.
• Measurement can distort behaviour: Optimizing clicks or time-on-platform may reduce wellbeing or customer value.

Entrepreneurs should therefore evaluate not only growth potential but also resilience, reversibility, social impact and the distribution of benefits and risks.

Contemporary Developments and Best Practices

Recent developments include generative AI assistants, composable cloud architectures, embedded finance, edge computing, digital identity, synthetic data and increasingly formal AI-risk management. These developments create opportunity but also increase the need for technical and institutional competence.

Best practices for entrepreneurs include:

1. Start with a customer or operational problem.
2. Treat trust, privacy and security as design requirements.
3. Use a narrow beachhead market to establish liquidity or product-market fit.
4. Build interoperable systems and document key dependencies.
5. Keep humans accountable for consequential decisions.
6. Test assumptions with measurable experiments.
7. Monitor unit economics as well as growth.
8. Establish data ownership, access, retention and deletion rules.
9. Maintain incident-response and business-continuity plans.
10. Review algorithms for drift, bias, explainability and unintended effects.
11. Design fair rules for platform participants, including appeals.
12. Plan for regulation, provider failure and changing technology costs.

Chapter Summary

Digital business models use technology to reshape value propositions, channels, operations and revenue mechanisms. Omnichannel models integrate physical and digital touchpoints into a coherent customer experience. Platforms create value by enabling interactions among participant groups, while ecosystems include the wider network of complements, infrastructure and institutions.

Network effects can accelerate adoption, but they do not guarantee profitability. Platforms must solve early coordination problems, maintain quality and govern participant behaviour. Cloud computing and automation can improve scalability and flexibility, but they introduce supplier, security and process risks. Data can support personalization, operational improvement and new services, yet its value depends on quality, lawful use and analytical capability.

AI can augment decisions, automate processes and become part of a product. Its successful use requires attention to data, workflow design, evaluation, human oversight and accountability. Cybersecurity, privacy and responsible AI are strategic elements of the business model rather than peripheral compliance activities. The strongest digital ventures combine customer value, sound economics, resilient technology and trustworthy governance.

Key Terms

Digital business model: A business model in which digital technology significantly shapes value creation, delivery or capture.

Omnichannel: An integrated approach connecting multiple customer and fulfilment channels.

Platform: An organization or system that enables interactions among two or more participant groups.

Ecosystem: A network of organizations, users, technologies and institutions jointly influencing value creation.

Network effect: A change in user value caused by a change in participation.

Direct network effect: A network effect arising among users of the same type.

Indirect network effect: A network effect transmitted between different participant groups.

Chicken-and-egg problem: The difficulty of attracting one platform side without the other.

Platform governance: Rules and mechanisms controlling participation and interactions.

Cloud computing: On-demand access to shared computing resources and services.

Automation: Use of technology to perform tasks with limited human intervention.

Data monetization: Creation of economic value from data, insights or data-enabled services.

Artificial intelligence: Computational capability supporting tasks such as prediction, classification, language processing or generation.

Machine learning: Methods that infer patterns from data to support predictions or decisions.

Cybersecurity: Protection of systems and information from unauthorized access, disruption or misuse.

Privacy: Appropriate control over the collection, use, disclosure and retention of personal information.

Responsible AI: Development and use of AI with attention to fairness, transparency, safety, privacy and accountability.

Customer acquisition cost: Average attributable sales and marketing cost per newly acquired customer.

Customer lifetime value: Estimated gross contribution generated by a customer over the relationship period.

Platform liquidity: The likelihood that participants can find and complete a valuable interaction.

Review Questions

5 Short Answer Questions

1. What distinguishes a platform business from a pipeline business?
2. Explain the difference between multichannel and omnichannel operations.
3. What is an indirect network effect?
4. State two benefits and two risks of cloud computing.
5. Why is human oversight important in consequential AI applications?

5 Long Answer Questions

1. Explain how a platform can overcome the chicken-and-egg problem during market entry.
2. Discuss the role of data quality, governance and privacy in data monetization.
3. Compare subscription, transaction-fee and advertising revenue models for digital ventures.
4. Explain how cybersecurity can affect a venture’s business model and reputation.
5. Discuss the benefits and limitations of AI-enabled automation from an entrepreneurial perspective.

3 Analytical / Application Questions

1. Design an omnichannel model for a small retailer. Identify the customer journey, technology requirements, operational risks and performance metrics.
2. A two-sided marketplace has many registered users but few completed transactions. Diagnose possible causes and recommend interventions involving pricing, trust, matching and governance.
3. Evaluate whether an AI system should be used to screen loan applications for a start-up lender. Address data quality, precision and recall, fairness, explainability, human review, privacy and accountability.

Multiple Choice Questions with Answers

Question 1

Which statement best describes a platform business model?

A. It transforms raw materials through a linear internal production process.

B. It enables interactions among two or more participant groups.

C. It sells only physical products through owned stores.

D. It avoids rules and governance to maximize participation.

Correct Answer: B. It enables interactions among two or more participant groups.

Question 2

Which example represents an indirect network effect?

A. More users of a messaging service make it more useful to each user.

B. More sellers attract buyers, while more buyers attract sellers.

C. A larger warehouse increases a firm’s storage capacity.

D. A faster processor reduces software execution time.

Correct Answer: B. More sellers attract buyers, while more buyers attract sellers.

Question 3

A company reports monthly ARPU of $30, a gross-margin ratio of 0.60 and monthly churn of 0.10. Using LTV = (ARPU × GM) ÷ c, what is the estimated LTV?

A. $18

B. $30

C. $180

D. $500

Correct Answer: C. $180

Question 4

What is the defining feature of an omnichannel model?

A. It uses only digital channels.

B. It operates several unrelated channels with separate customer records.

C. It integrates channels into a connected customer and operating experience.

D. It eliminates physical fulfilment.

Correct Answer: C. It integrates channels into a connected customer and operating experience.

Question 5

Which cloud service category provides complete applications to end users?

A. Infrastructure as a Service

B. Platform as a Service

C. Software as a Service

D. Network as a Service

Correct Answer: C. Software as a Service

Question 6

If TP = 80 and FP = 20, what is precision?

A. 20%

B. 50%

C. 80%

D. 100%

Correct Answer: C. 80%

Question 7

Which control most directly reduces the risk of account takeover?

A. Multi-factor authentication

B. Increasing advertising expenditure

C. Removing all system logs

D. Publishing raw customer data

Correct Answer: A. Multi-factor authentication

Question 8

Which is an example of data monetization through an insight product?

A. Deleting all operational records after collection

B. Selling a forecast based on aggregated and analyzed demand data

C. Using a password manager internally

D. Replacing a website logo

Correct Answer: B. Selling a forecast based on aggregated and analyzed demand data

Question 9

Why can a large number of platform registrations be a misleading success measure?

A. Registrations always increase cybersecurity risk.

B. Registered users may not engage in successful or repeated transactions.

C. Platforms cannot measure registrations accurately.

D. Registrations are illegal in most markets.

Correct Answer: B. Registered users may not engage in successful or repeated transactions.

Question 10

Which principle is most closely associated with responsible AI?

A. Maximizing automation regardless of consequences

B. Avoiding all human intervention

C. Ensuring accountability, appropriate oversight and fair treatment

D. Keeping every algorithm secret from its operators

Correct Answer: C. Ensuring accountability, appropriate oversight and fair treatment