Managing AI Innovation and Technology Strategy

Introduction

Artificial intelligence (AI) is no longer only a technical capability located in an organization’s research laboratory. In media organizations it influences audience analytics, content recommendation, automated transcription, translation, advertising optimization, newsroom workflows, production, rights management, customer service, and strategic decision-making. Yet acquiring an AI tool does not automatically create organizational value. Value emerges when a media organization identifies a meaningful problem, selects an appropriate technology, redesigns work, manages risk, develops staff capability, and measures outcomes.

AI innovation management is therefore a strategic and organizational discipline. It connects business objectives with technology choices and asks five central questions:

1. What problem should AI solve?
2. Is the organization ready to adopt it responsibly?
3. Which technological and vendor option is most appropriate?
4. What will the initiative cost, and what benefits are realistic?
5. How will the organization implement, govern, evaluate, and improve it?

For media managers, these questions are especially important because media products depend on trust, creativity, intellectual property, speed, public accountability, and relationships with audiences and advertisers. A technically impressive system may still fail if it damages editorial credibility, increases hidden labour, violates rights, or produces no sustainable economic benefit.

Learning Objectives

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

• construct a business case for an AI-enabled media initiative;
• assess organizational readiness across strategy, data, people, processes, technology, and governance;
• compare vendors and technology options using transparent criteria;
• estimate costs, benefits, return on investment, and payback period;
• plan AI projects using staged delivery and appropriate change-management practices;
• identify workforce capabilities required for responsible adoption;
• design meaningful key performance indicators (KPIs); and
• evaluate ethical, legal, operational, and strategic risks.

1. From Technology Enthusiasm to Strategic Value

Technology strategy is the deliberate alignment of technology investments with organizational purpose and competitive priorities. AI strategy is a subset of technology strategy concerned with systems that perform tasks commonly associated with human intelligence, such as classification, prediction, language generation, speech recognition, recommendation, and pattern detection.

An AI initiative should be treated as a portfolio of organizational change rather than as a software purchase. Its components may include:

• a business or editorial problem;
• data and data-governance arrangements;
• models, applications, and infrastructure;
• redesigned workflows and decision rights;
• staff training and support;
• risk controls and accountability; and
• measurement of benefits and unintended effects.

A useful strategic distinction is between automation, augmentation, and innovation. Automation transfers a task from a person to a system. Augmentation enables people to perform an existing task more effectively. Innovation creates a new product, service, market relationship, or operating model. For example, automatic captioning may automate transcription, an editorial research assistant may augment journalists, and a personalized audio-news service may represent product innovation.

These categories are not mutually exclusive. A single initiative can automate routine work while augmenting professional judgement and enabling a new product. However, the distinction helps managers identify the appropriate success criteria. Automation may be assessed through time and error reduction; augmentation through quality and employee experience; innovation through audience growth, revenue, or strategic differentiation.

The resource-based and dynamic-capabilities perspectives

The resource-based view argues that sustainable advantage depends on valuable, rare, difficult-to-imitate, and organizationally supported resources. AI software itself is often widely available and therefore not necessarily a durable source of advantage. Differentiation may instead arise from proprietary audience data, trusted brands, distinctive archives, skilled employees, editorial processes, and the ability to integrate technology into daily operations.

The dynamic-capabilities perspective adds that organizations must sense opportunities, seize them through investment, and transform routines as conditions change. In a media company, this may involve detecting a change in audience behaviour, piloting an AI recommendation service, and then restructuring content planning and marketing around evidence from the pilot.

The implication is important: purchasing the same model or platform as competitors rarely guarantees competitive advantage. The strategic asset is the organization’s ability to apply technology responsibly and repeatedly.

2. Building an AI Business Case

A business case explains why an initiative should be funded, how it will create value, what resources it requires, and what risks decision-makers accept. It should not begin with a preferred vendor or a fashionable technology. It should begin with a clearly defined problem.

Problem definition

A strong problem statement identifies:

• the affected users or stakeholders;
• the current process and its weaknesses;
• the scale and frequency of the problem;
• the consequences of not acting;
• the desired future state; and
• constraints such as budget, regulation, rights, or editorial policy.

“Use generative AI in the newsroom” is not a business problem. “Reduce the time required to produce accurate first-pass transcripts for regional video reports while retaining human editorial review” is more useful because it identifies a process, a measurable objective, and a control requirement.

Alternatives analysis

The business case should compare at least three options:

1. continue with the current process;
2. improve the process without AI; and
3. adopt an AI-enabled process.

This prevents technology bias. A workflow redesign, better search interface, additional staff, or improved metadata may solve the problem more effectively than AI.

Value categories

Benefits may be financial, operational, editorial, audience-related, or strategic. Examples include:

Managers should distinguish cashable benefits from capacity benefits. A reduction in staff time does not necessarily reduce payroll expenditure. It may instead allow employees to produce more content, improve quality, or undertake higher-value work. The business case must state which interpretation is intended.

Financial evaluation

For a simple initiative, total cost of ownership can be represented as:

Here, C₀ is the initial acquisition or implementation cost; Cₜ is operating cost in period t; Mₜ is maintenance or integration cost; Trₜ is training and change-support cost; Gₜ is governance, monitoring, and assurance cost; and n is the number of periods. The formula assumes that costs are measured consistently, such as in pounds or dollars, and that undiscounted totals are acceptable for an initial estimate.

A basic return on investment is:

where B is the total measurable benefit and C is total cost over the same period. ROI is a simple ratio, not a complete investment appraisal. It does not show when benefits occur, how uncertain they are, or whether alternative investments would perform better.

Payback period estimates how long it takes for cumulative benefits to recover the initial investment. If benefits and costs are approximately constant, then:

This shortcut applies only when periodic net benefits are reasonably stable. If they vary, cumulative cash flows should be listed period by period.

For longer-term decisions, discounted cash flow may be used:

CFₜ is net cash flow in period t, r is the discount rate per period, and n is the number of periods. A positive net present value (NPV) indicates that discounted benefits exceed discounted costs under the selected assumptions. The method assumes that cash flows and the discount rate are estimated consistently and that the project’s risks are reflected adequately in the rate or in scenario analysis.

#### Worked example: transcription service

A broadcaster considers an AI transcription service. Initial integration costs are $24,000. Training and workflow redesign cost $6,000. The service costs $1,500 per month. It is expected to create $5,000 of measurable monthly benefit through reduced outsourcing and increased production capacity.

For the first year:

• Initial cost = $24,000 + $6,000 = $30,000.
• Annual operating cost = 12 × $1,500 = $18,000.
• Total first-year cost = $48,000.
• Annual benefit = 12 × $5,000 = $60,000.
• Net benefit = $60,000 − $48,000 = $12,000.
• ROI = ($12,000 ÷ $48,000) × 100% = 25%.

The simple payback calculation based on the initial implementation cost and monthly net benefit is also informative. Monthly net benefit is $5,000 − $1,500 = $3,500, so payback is approximately $30,000 ÷ $3,500 = 8.57 months. This does not mean the initiative is risk-free. Accuracy variation, contract escalation, data-protection costs, employee acceptance, and rights issues should be tested through scenarios.

3. Assessing Organizational Readiness

Readiness is the extent to which an organization has the conditions required to adopt and sustain a technology. It is not equivalent to technical maturity. A company may possess modern infrastructure but lack clear ownership, reliable data, or staff confidence.

A practical readiness assessment examines six dimensions:

A simple scoring model can support discussion. Let each dimension receive a score from 1 to 5, where 1 means severely inadequate and 5 means strong readiness. If all dimensions are weighted equally, the mean readiness score is:

Rᵢ is the score for dimension i, and k is the number of dimensions. The arithmetic mean should not be interpreted as a probability or as an objective scientific measurement. A low score in a critical dimension may be a gating issue even when the overall average is high. Organizations may therefore use minimum thresholds, such as requiring data and governance scores of at least 3 before a pilot.

Readiness assessments should include interviews, document review, workflow observation, data-quality checks, and technical discovery. They should also consider organizational culture. Employees may resist AI because of job insecurity, previous failed projects, inadequate consultation, or concern that professional standards will be weakened. Such concerns are evidence about implementation conditions, not merely obstacles to be removed.

4. Selecting Vendors and Technology Options

Vendor selection should be based on requirements rather than demonstrations. A polished demonstration may conceal limitations involving data retention, model drift, integration, service outages, or intellectual property.

Requirements specification

Requirements should be divided into:

• functional requirements: what the system must do;
• quality requirements: accuracy, latency, availability, accessibility, and usability;
• governance requirements: audit logs, human review, explainability, privacy, and rights management;
• commercial requirements: pricing, contract duration, support, exit provisions, and service levels; and
• integration requirements: application programming interfaces, identity management, content-management systems, archives, and analytics platforms.

Weighted scoring

A weighted decision matrix makes evaluation explicit. Suppose criteria have weights that sum to 1 and vendors receive scores from 1 to 5. The weighted score is:

where Sⱼ is the total score for vendor j, wᵢ is the weight of criterion i, and xᵢⱼ is vendor j’s score on that criterion. The score assumes that criteria are defined consistently and that higher scores always represent better performance. If a criterion is a risk, such as vendor lock-in, it should be scored so that a lower risk receives a higher score.

For example, a media organization may assign the following weights: editorial quality 0.30, security and privacy 0.25, integration 0.20, cost 0.15, and support 0.10. Vendor scores should be supported by tests, references, contract review, and evidence—not by sales claims alone.

Important due-diligence questions include:

• Are customer prompts, content, or metadata retained for model training?
• Where are data stored and processed?
• What rights does the vendor require over inputs and outputs?
• How are errors, incidents, and service outages reported?
• Can the organization export its data and migrate away?
• What happens if the vendor changes the model or pricing?
• Are performance claims valid for the organization’s languages, accents, genres, and audience groups?
• Does the contract permit independent audit and termination for material risk?

A pilot should test real, representative workloads. Synthetic demonstrations are useful for exploration but insufficient for procurement decisions.

5. Project Delivery and Change Management

AI projects are socio-technical projects: technical design and human work are inseparable. A responsible delivery lifecycle normally includes discovery, feasibility, pilot, controlled rollout, and continuous evaluation.

Discovery

The team maps the current workflow, defines the problem, identifies stakeholders, establishes baseline measures, and records assumptions. Stakeholders may include journalists, producers, editors, data-protection officers, legal advisers, engineers, accessibility specialists, unions, audiences, and vendors.

Pilot

The pilot should be limited in scope but realistic in context. It should specify success thresholds, human-review requirements, incident procedures, and a stopping rule. A stopping rule defines conditions under which the project will pause or terminate, such as unacceptable error rates, unresolved rights concerns, or failure to achieve minimum productivity gains.

Rollout

Scaling requires integration into systems, training, documentation, support, monitoring, and governance. The organization should avoid treating launch as the end of the project. AI performance may change when data, audiences, models, or editorial practices change.

Change management should address four questions:

1. What will change in people’s daily work?
2. Why is the change necessary and what benefits are expected?
3. What authority and accountability will remain with human professionals?
4. What learning, support, and feedback mechanisms are available?

Communication should be two-way. Staff need opportunities to report failures and propose improvements. Managers should not reward rapid adoption at the expense of careful use. Incentives must reinforce quality, compliance, and public trust.

6. Workforce Capability and Operating Models

AI adoption changes the skills required in media organizations. Technical specialists are important, but responsible adoption also depends on domain professionals who understand editorial standards, audience needs, and cultural context.

A capability framework may include:

• AI literacy: understanding capabilities, limitations, uncertainty, and common failure modes;
• data literacy: interpreting data quality, bias, metadata, and measurement;
• prompt and workflow design: specifying tasks, constraints, review steps, and reusable procedures;
• editorial judgement: verifying facts, sources, tone, context, and public-interest implications;
• technical integration: APIs, identity, security, testing, monitoring, and data pipelines;
• legal and governance competence: privacy, copyright, licensing, accessibility, records, and accountability;
• product and business analysis: defining users, value propositions, costs, benefits, and experiments; and
• leadership capability: making decisions under uncertainty and coordinating multidisciplinary teams.

Organizations should distinguish training from capability development. A one-hour demonstration may create awareness, but sustained competence requires practice, feedback, documentation, peer learning, and access to expert support. Role-based training is usually more effective than a single generic course.

Possible operating models include a centralized AI team, embedded specialists in departments, or a hub-and-spoke model combining central standards with local experimentation. Centralization may improve governance and technical efficiency; local teams may understand workflows and audience needs better. The appropriate model depends on organizational size, risk, and strategic priorities.

7. KPIs, Evaluation, and Learning

KPIs should connect technology activity to organizational outcomes. Counting the number of AI tools purchased or prompts submitted measures activity, not value.

A balanced measurement framework may include:

Each KPI requires a definition, data source, owner, frequency, baseline, target, and action threshold. For example, “quality improved” is weak. “The proportion of transcripts requiring major correction falls from 18% to below 10% over six months, measured through a defined editorial sample” is more operational.

Evaluation should use baselines and, where feasible, comparison groups. If a new recommendation system is introduced, audience changes may result from seasonality, marketing, or content changes rather than the system itself. A controlled experiment can help, but it must respect consent, fairness, editorial obligations, and user experience. Quantitative indicators should be supplemented by qualitative review, especially when outcomes involve trust, creativity, or public value.

AI systems also require monitoring for drift. Drift may occur when input data, language, audience behaviour, content genres, or model versions change. Monitoring should track performance across relevant groups rather than relying only on an overall average.

8. Responsible Governance and Risk Management

AI governance allocates responsibility for decisions, controls, monitoring, and remediation. It should be proportionate to risk. A system that recommends archive keywords presents different risks from one that determines access to employment, generates public-news content, or moderates user speech.

A risk register may record:

• risk description and affected stakeholders;
• likelihood and impact assessment;
• existing controls;
• responsible owner;
• residual risk;
• indicators and review date; and
• response, such as avoid, mitigate, transfer, or accept.

Major media risks include:

• inaccuracy and hallucination: generated material may appear plausible but be false;
• bias and exclusion: systems may perform poorly for languages, dialects, communities, or disability groups;
• privacy: personal data may be collected, inferred, exposed, or transferred unlawfully;
• copyright and licensing: training data, inputs, outputs, and archival material may involve rights restrictions;
• security: prompt injection, data leakage, account compromise, and supply-chain vulnerabilities;
• opacity: users may not understand how a recommendation or classification was produced;
• labour and professional identity: automation may intensify work, reduce autonomy, or displace roles; and
• trust and authenticity: audiences may be misled about whether content is human-created, AI-assisted, or synthetic.

Controls can include human review, restricted data access, logging, provenance records, disclosure labels, model testing, red-teaming, incident response, contractual safeguards, and periodic independent review. Human oversight must be meaningful: a reviewer needs sufficient time, competence, authority, and access to evidence to challenge the system.

Legal obligations vary by jurisdiction and application. Managers should consult applicable privacy, copyright, consumer-protection, employment, accessibility, media, and AI-specific laws. Compliance is a minimum requirement, not a substitute for ethical judgement or public accountability.

9. Applied Case: Northbridge Public Media

This is an illustrative fictional case. Northbridge Public Media (NPM) operates television, radio, web, and podcast services in several languages. Its accessibility team spends substantial time correcting automated captions and manually creating transcripts. Senior management proposes purchasing a generative AI platform that promises faster transcription, translation, summarization, and social-media copy.

Background and problem

The current process is accurate but slow. Online clips are often published many hours after broadcast because transcripts and captions are prepared manually. The organization wants faster publication without reducing accessibility or editorial reliability. Staff are concerned that the proposed platform may retain sensitive interview content and may perform poorly with regional accents.

Analysis

The project team defines the objective as: reduce average transcript preparation time by 40% within six months while maintaining a major-error rate below the existing baseline and ensuring that no unreviewed transcript is published.

The team maps the workflow and establishes a baseline from 200 recent programmes. It identifies three alternatives: additional freelance transcription, a non-AI workflow redesign using improved templates, and an AI-assisted process. It also identifies non-financial benefits, including faster access for audiences with hearing impairments and improved searchability of the archive.

Readiness assessment shows strong strategic alignment and staff expertise, but weak vendor-risk documentation and uncertain data-retention arrangements. The project is therefore not approved for immediate full deployment.

Approach and solution

NPM issues a requirements specification. Vendors must provide data-location information, deletion controls, contractual restrictions on model training, performance tests using NPM’s languages and accents, audit logs, export capability, and service-level commitments.

A six-week pilot is conducted on a representative sample. The AI produces first-pass transcripts, but trained staff verify names, numbers, quotations, and sensitive content. The team records editing time, major and minor errors, accessibility measures, employee confidence, and incidents. A clear label identifies AI-assisted material internally, and publication remains subject to editorial approval.

The pilot shows a 35% reduction in average preparation time—below the target—but a larger gain for routine studio programmes than for live regional interviews. The organization therefore adopts the system only for selected programme types, retains manual transcription for high-risk content, and funds additional language-specific testing. The contract includes an exit clause and prohibits vendor use of NPM content for general model training without explicit authorization.

Outcome and learning points

NPM does not define success as replacing transcription staff. Staff time is redirected toward quality assurance, accessibility improvement, archive enrichment, and complex interviews. The project produces moderate operational benefit and significant learning about data, workflow variation, and language performance.

The case demonstrates that:

• a precise problem is more useful than a broad technology ambition;
• alternatives should be compared before procurement;
• readiness weaknesses can be corrected before scaling;
• representative testing is essential;
• human review must be designed into the workflow; and
• partial adoption may be strategically better than universal deployment.

Contemporary Developments

Several developments are reshaping AI technology strategy in media. Generative AI is moving from isolated experimentation toward integrated workflow tools, but reliability and rights management remain unresolved. Multimodal systems can process text, audio, images, and video, creating opportunities for archive search and production while increasing the complexity of evaluation. Smaller or specialized models may offer lower cost, greater control, and better performance for particular languages or domains. Open-source models may reduce dependence on a single vendor but increase internal responsibilities for security, infrastructure, and maintenance.

Synthetic media and digital provenance are also becoming strategic concerns. Organizations increasingly need processes that record how content was created, edited, and approved. Energy use and infrastructure impacts may influence procurement decisions, especially for large-scale model training or high-volume inference. Finally, regulation and professional standards are evolving, making legal monitoring and policy review continuing management responsibilities rather than one-time tasks.

Best-Practice Principles

1. Start with a measurable organizational problem, not a technology trend.
2. Compare AI with non-AI alternatives.
3. Treat data, workflow, people, and governance as part of the solution.
4. Use representative pilots with explicit success and stopping criteria.
5. Evaluate total cost of ownership rather than headline subscription price.
6. Make vendor claims testable through evidence and contract terms.
7. Preserve meaningful human accountability for consequential decisions.
8. Measure quality, trust, equity, and public value alongside productivity.
9. Invest in role-based capability development and employee participation.
10. Plan for monitoring, model change, incidents, and exit from the beginning.

Chapter Summary

Managing AI innovation requires alignment between technology and organizational purpose. A credible AI business case begins with a defined problem, considers non-AI alternatives, identifies measurable benefits, estimates total cost of ownership, and makes assumptions explicit. ROI, payback period, and NPV can support decisions, but they must be interpreted alongside uncertainty, public value, and risk.

Organizational readiness depends on strategy, data, processes, technology, people, and governance. Vendor selection should use requirements, representative testing, weighted criteria, due diligence, and contract safeguards. AI projects are socio-technical projects requiring discovery, piloting, staged rollout, communication, training, and continuous monitoring.

Sustainable adoption depends on workforce capability and meaningful KPIs. Responsible governance addresses accuracy, bias, privacy, copyright, security, labour, transparency, and audience trust. The central strategic lesson is that AI creates value not simply because it is powerful, but because an organization can apply it deliberately, responsibly, and measurably.

Key Terms

AI business case: A structured justification for investing in an AI initiative.

Artificial intelligence: Computational systems that perform tasks associated with human intelligence.

Automation: Transfer of a task from human performance to a system.

Augmentation: Use of technology to enhance human capability and judgement.

Business benefit: A measurable positive outcome produced by an initiative.

Change management: Practices that support adoption of new technologies, processes, and roles.

Dynamic capabilities: Organizational abilities to sense opportunities, seize them, and transform resources.

Governance: Allocation of authority, accountability, rules, and controls.

Human oversight: Meaningful human ability to review, challenge, and intervene in system outputs.

Key performance indicator: A defined measure used to monitor progress toward an objective.

Net present value: The value of future net cash flows after discounting them to the present.

Organizational readiness: The extent to which an organization has conditions necessary for adoption.

Payback period: Time required for cumulative benefits to recover an investment.

Pilot: A limited, controlled implementation used to test feasibility and value.

Resource-based view: A theory that competitive advantage may arise from valuable and difficult-to-imitate organizational resources.

Return on investment: Net benefit divided by cost, expressed as a percentage.

Total cost of ownership: The complete acquisition, operation, maintenance, training, and governance cost of a system.

Vendor due diligence: Investigation of a supplier’s capabilities, risks, contracts, security, and performance.

Workflow redesign: Deliberate restructuring of activities, roles, information, and decision points.

Review Questions

Short Answer Questions

1. Distinguish between automation, augmentation, and innovation in an AI media initiative.
2. What six dimensions should be examined in an organizational readiness assessment?
3. Why should a business case compare AI with non-AI alternatives?
4. State two important vendor due-diligence questions for a media organization.
5. Why is counting the number of AI tools purchased an inadequate KPI?

Long Answer Questions

1. Explain how a media organization can construct a robust AI business case, including benefits, costs, assumptions, and risks.
2. Discuss the resource-based view and dynamic-capabilities perspective in relation to AI strategy.
3. Explain how a weighted vendor-selection matrix can improve procurement decisions and identify its limitations.
4. Discuss the role of workforce capability and change management in responsible AI adoption.
5. Evaluate the governance issues raised by AI-assisted content production in media organizations.

Analytical / Application Questions

1. A streaming service is considering an AI recommendation system. Design a readiness assessment covering data, people, process, technology, strategy, and governance, and identify two possible gating conditions.
2. An AI tool costs $18,000 initially, $2,000 for training, and $1,200 per month. It is expected to generate $4,000 of monthly measurable benefit. Calculate first-year total cost, first-year net benefit, and simple ROI. State two assumptions behind your calculation.
3. Develop a KPI framework for an AI-assisted newsroom. Include at least one financial, operational, quality, workforce, audience, and governance measure, and explain how each should be interpreted.

Multiple Choice Questions with Answers

Question 1

Which statement best describes a strong AI business case?

A. It begins with a popular vendor and seeks a problem later.

B. It links a defined problem to benefits, costs, risks, and alternatives.

C. It focuses only on technical performance.

D. It assumes all staff time savings become cash savings.

Correct Answer: B. It links a defined problem to benefits, costs, risks, and alternatives.

Question 2

Which of the following is a readiness dimension?

A. Logo design

B. Office decoration

C. Data governance

D. Advertising slogan length

Correct Answer: C. Data governance

Question 3

A project has total measurable benefits of $80,000 and total costs of $50,000. What is its simple ROI?

A. 30%

B. 37.5%

C. 60%

D. 160%

Correct Answer: C. 60%

Question 4

What is the main purpose of a pilot?

A. To guarantee that a system will never fail

B. To test feasibility and value under controlled, realistic conditions

C. To avoid defining success criteria

D. To replace procurement due diligence

Correct Answer: B. To test feasibility and value under controlled, realistic conditions

Question 5

In a weighted vendor matrix, what do the weights represent?

A. The number of employees using each vendor

B. The relative importance of evaluation criteria

C. The vendor’s annual profit

D. The system’s processing speed only

Correct Answer: B. The relative importance of evaluation criteria

Question 6

Which benefit is not necessarily cashable?

A. A cancelled software invoice

B. A reduction in external contractor payments

C. Employee time released for additional higher-value work

D. A contracted reduction in subscription fees

Correct Answer: C. Employee time released for additional higher-value work

Question 7

Which control most directly supports meaningful human oversight?

A. Removing all review stages

B. Allowing publication without verification

C. Giving trained reviewers authority and time to challenge outputs

D. Hiding system limitations from users

Correct Answer: C. Giving trained reviewers authority and time to challenge outputs

Question 8

Why is total cost of ownership important?

A. It excludes training and governance costs.

B. It considers only the initial purchase price.

C. It captures acquisition, operation, maintenance, training, and governance costs.

D. It guarantees a positive ROI.

Correct Answer: C. It captures acquisition, operation, maintenance, training, and governance costs.

Question 9

Which KPI best measures content-quality risk?

A. Number of AI demonstrations attended

B. Number of prompts submitted

C. Major factual error rate after editorial review

D. Number of vendor brochures received

Correct Answer: C. Major factual error rate after editorial review

Question 10

A system performs well overall but poorly for a particular language group. What issue does this most clearly indicate?

A. Potential bias or unequal performance

B. Successful universal deployment

C. Lower total cost of ownership

D. Improved vendor portability

Correct Answer: A. Potential bias or unequal performance