I85PREP
⌘ K
CAIM study materials
CAIM study resources

CAIM practice questions

One question at a time. Build your CAIM understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1AI Strategy and Governance

An organization is drafting its first AI governance charter. Which element most directly establishes accountability for AI outcomes across the enterprise?

  1. A procurement checklist limited to hardware vendors
  2. A named executive owner for AI risk with defined escalation paths Correct answer.
  3. A curated library of approved open-source model weights
  4. A quarterly newsletter summarizing AI industry trends

Explanation

Governance charters create accountability by assigning explicit ownership and decision rights, so a named executive owner with escalation paths is the core accountability mechanism. The other items are operational or informational and do not allocate responsibility for AI outcomes.

Question 2AI Strategy and Governance

A bank wants its AI governance body to remain effective as models proliferate. Which design choice best supports scalable oversight?

  1. Delegating all AI decisions to individual data scientists
  2. Requiring the full board to review every model change
  3. Tiering AI use cases by risk so review depth matches impact Correct answer.
  4. Freezing all AI deployments until a new policy is written

Explanation

Risk-tiered review lets governance bodies concentrate scrutiny on high-impact systems while standardizing lighter review for low-risk uses, which scales as the model inventory grows. Board-level review of every change and full delegation both break proportionality.

Question 3AI Strategy and Governance

Which artifact most clearly documents the intended use and known limitations of a deployed AI system for downstream reviewers?

  1. A source code commit log
  2. A cloud cost dashboard
  3. A model card or system fact sheet Correct answer.
  4. A vendor's marketing brochure

Explanation

Model cards and system fact sheets are the standard transparency artifacts describing intended use, performance, and limitations for reviewers and users. Commit logs, cost dashboards, and brochures do not convey scope and limitation information in a reviewable form.

Question 4AI Strategy and Governance

An AI strategy team must align model deployment with existing data protection obligations. Which governance control most directly enforces that alignment before launch?

  1. An annual employee satisfaction survey
  2. A monthly infrastructure capacity report
  3. A post-incident root cause review
  4. A privacy impact assessment gate in the release process Correct answer.

Explanation

Embedding a privacy impact assessment as a release gate forces the team to evaluate data protection obligations before deployment, which is a preventive control. Post-incident reviews and capacity reports are detective or operational and occur outside the compliance decision point.

Question 5AI Strategy and Governance

A governance committee is defining escalation criteria for AI incidents. Which criterion best distinguishes an incident requiring executive notification from one handled by the operations team?

  1. The severity of potential harm to individuals or the organization Correct answer.
  2. The number of engineers who touched the model code
  3. The age of the underlying training dataset
  4. The programming language used in the pipeline

Explanation

Incident escalation should be driven by potential impact, so severity of harm to people or the organization is the appropriate threshold for executive notification. Technical implementation details and dataset age do not measure the consequences that justify escalation.

Question 6AI Strategy and Governance

Which practice best supports meaningful human oversight of a high-risk AI decision system?

  1. Reviewers approve decisions in batches under time pressure
  2. Reviewers are instructed to defer to the model in all cases
  3. Reviewers see only the final model output with no context
  4. Reviewers receive decision rationale and authority to override the model Correct answer.

Explanation

Effective human oversight requires that reviewers understand the basis of a recommendation and can act on it, including overriding the system. Reviewers without rationale, under extreme time pressure, or told to defer cannot exercise genuine judgment.

Question 7AI Strategy and Governance

An organization wants to measure whether its AI governance program is functioning, not just documented. Which indicator is most diagnostic of program effectiveness?

  1. The number of AI conferences employees attended
  2. The percentage of high-risk models that completed required reviews before release Correct answer.
  3. The total count of AI-related job postings
  4. The number of pages in the AI policy document

Explanation

Governance effectiveness is shown by whether required controls are actually applied, so the share of high-risk models that completed mandated reviews before release is a strong diagnostic. Policy length, conference attendance, and hiring volume measure activity or documentation rather than control adherence.

Question 8AI Strategy and Governance

A company plans to use a third-party foundation model for customer-facing advice. Which governance action best addresses the risk that the vendor changes the model without notice?

  1. Assume model behavior is stable once initially validated
  2. Remove internal testing to reduce duplicated effort
  3. Negotiate contractual notice and change-management commitments from the vendor Correct answer.
  4. Rely on the vendor's public status page for outage alerts

Explanation

Third-party model risk is mitigated through contract terms that require advance notice and defined change management, giving the organization time to retest. Status pages cover availability, not model behavior changes, and assuming stability or cutting internal testing increases exposure.

Question 9AI Strategy and Governance

Which statement best describes the purpose of an AI use case inventory in a governance program?

  1. It replaces the need for risk assessment by listing systems
  2. It guarantees that no unapproved AI tools exist in the organization
  3. It provides visibility into where AI is used so risk can be assessed and monitored Correct answer.
  4. It documents only models built by the internal data science team

Explanation

An inventory creates the visibility needed to identify, prioritize, and monitor AI uses across the organization; it is an input to risk assessment, not a substitute for it. It also cannot guarantee complete coverage or be limited to internally built models without creating blind spots.

Question 10AI Strategy and Governance

A model used for credit decisions shows higher error rates for one demographic group. Which governance response is most appropriate at the strategy level?

  1. Remove the demographic field from reporting to simplify metrics
  2. Deploy the model and monitor complaints after launch
  3. Halt deployment pending bias remediation and documented review Correct answer.
  4. Increase the model's decision threshold uniformly for all applicants

Explanation

A material disparity in error rates for a high-stakes decision is a governance red flag that warrants pausing deployment until the disparity is remediated and reviewed. Deploying with after-the-fact monitoring, hiding the metric, or applying a uniform threshold does not address the identified harm.

Question 11AI Strategy and Governance

Which role is typically responsible for independently challenging an AI risk assessment before a high-risk system is approved?

  1. The model's original developer
  2. The marketing team that requested the feature
  3. The vendor that supplied the training data
  4. A second-line risk or compliance function Correct answer.

Explanation

Independent challenge is a hallmark of second-line risk and compliance functions, which are separated from the teams building and requesting the system. Developers, vendors, and requesting business units have conflicts of interest that undermine independent review.

Question 12AI Strategy and Governance

An organization wants its AI principles to influence day-to-day engineering decisions rather than remain aspirational. Which approach best achieves that?

  1. Require engineers to memorize the principles verbatim
  2. Publish the principles on the corporate website only
  3. Restrict principles discussions to the legal department
  4. Translate principles into concrete engineering checklists and review criteria Correct answer.

Explanation

Principles become operational when they are converted into specific checks, templates, and review criteria that engineers apply in their workflow. Publication alone, rote memorization, or confining the discussion to legal keeps principles disconnected from engineering practice.

Question 13AI Strategy and Governance

Which consideration most strongly supports establishing a formal AI ethics review board rather than relying on ad hoc consultation?

  1. It transfers all liability to individual board members
  2. It reduces the need for any technical testing
  3. It guarantees that no ethical disputes will arise
  4. It creates a consistent, documented forum for evaluating contested use cases Correct answer.

Explanation

A standing review board provides a repeatable, documented process for deliberating difficult cases and recording decisions, which ad hoc consultation lacks. It does not replace technical testing, prevent all disputes, or shift liability to individuals.

Question 14AI Strategy and Governance

A team proposes tracking model drift as part of AI governance monitoring. Why is drift relevant to governance rather than only to engineering?

  1. Drift can degrade performance and change the risk profile of a deployed system Correct answer.
  2. Drift is relevant only to models trained on synthetic data
  3. Drift only affects the cost of cloud storage
  4. Drift eliminates the need for periodic model review

Explanation

Drift can cause a model's behavior and error rates to change after deployment, which alters the risk profile and may invalidate prior approvals. That makes drift monitoring a governance concern, not merely a technical tuning issue.

Question 15AI Strategy and Governance

Which element is essential for an AI audit trail to be useful during a post-incident investigation?

  1. A summary of the team's preferred coding standards
  2. The personal opinions of the product manager
  3. Records of model versions, inputs, and decisions with timestamps Correct answer.
  4. A list of competitor AI products

Explanation

Investigations require reconstructing what the system did and why, which depends on versioned records of inputs, outputs, and decisions with timestamps. Coding preferences, personal opinions, and competitor lists do not support reconstruction of system behavior.

Question 16AI Strategy and Governance

An organization wants to ensure AI governance keeps pace with rapidly evolving models. Which structural choice best supports that goal?

  1. An informal agreement among engineering leads
  2. A one-time policy approved at program launch
  3. A scheduled review cycle with triggers for significant model or regulatory change Correct answer.
  4. A policy owned solely by the IT helpdesk

Explanation

Governance stays current when policies are reviewed on a cadence and re-examined when material changes occur, such as new model capabilities or regulations. One-time policies, misplaced ownership, and informal agreements lack the mechanism to adapt.

Question 17AI Strategy and Governance

Which practice best supports responsible data sourcing for an AI training program under a governance framework?

  1. Deleting provenance records after training completes
  2. Relying on the assumption that public data is unrestricted
  3. Collecting the largest possible dataset regardless of source
  4. Documenting data provenance, licensing, and permitted uses Correct answer.

Explanation

Responsible sourcing depends on knowing where data came from, what rights accompany it, and whether the intended use is permitted, which requires documented provenance and licensing. Maximizing volume, assuming public data is unrestricted, or discarding records undermines that control.

Question 18AI Strategy and Governance

A regulator asks an organization to demonstrate that its AI systems are subject to ongoing risk management. Which evidence most directly supports that claim?

  1. A signed executive memo declaring AI is important
  2. An org chart showing the data science team
  3. Periodic risk assessments with documented remediation tracking Correct answer.
  4. A list of AI vendors the organization evaluated

Explanation

Ongoing risk management is evidenced by repeated assessments and tracked remediation, showing the process operates over time. A declaration, vendor list, or org chart shows intent or structure but not sustained risk management activity.

Question 19AI Strategy and Governance

Which factor most appropriately raises the governance intensity applied to an AI use case?

  1. The model uses a popular open-source framework
  2. The development team is geographically distributed
  3. The use case affects access to essential services or safety Correct answer.
  4. The model was trained in under a day

Explanation

Governance intensity should scale with potential impact, and use cases affecting essential services or safety carry higher stakes. Framework popularity, team distribution, and training speed are implementation facts that do not determine impact.

Question 20AI Strategy and Governance

Which statement best captures the relationship between an AI strategy and an AI governance framework?

  1. They are unrelated and owned by separate organizations
  2. Governance applies only to failed AI projects
  3. Governance replaces strategy once a policy is approved
  4. Strategy sets direction and value goals while governance sets guardrails and accountability Correct answer.

Explanation

Strategy defines where the organization is going with AI and why, while governance defines the boundaries, controls, and accountability that keep that pursuit safe and lawful. They are complementary and both apply across the AI portfolio.

Question 21AI Strategy and Governance

An executive asks why AI governance should be funded before a major incident occurs. Which argument is most persuasive?

  1. Governance spending guarantees that models will never fail
  2. Governance funding is required only for marketing purposes
  3. Preventive controls reduce the likelihood and cost of harm compared with remediation after failure Correct answer.
  4. Incidents are unrelated to the presence or absence of controls

Explanation

The business case for proactive governance rests on reducing the probability and severity of harm, which lowers remediation, legal, and reputational costs. No control set guarantees zero failures, and claiming incidents are unrelated to controls contradicts the evidence.

Question 22AI Strategy and Governance

Which approach best addresses the risk of shadow AI use by employees adopting unapproved tools?

  1. Banning all AI tools without offering alternatives
  2. Providing an approved tool path with clear usage guidance and detection Correct answer.
  3. Relying solely on an annual attestation
  4. Ignoring unapproved use to avoid conflict

Explanation

Shadow AI is best reduced by giving employees a sanctioned, easy-to-use path with guidance while monitoring for unapproved use. Blanket bans without alternatives drive usage further underground, and ignoring the issue or relying on annual attestations leaves the risk unmanaged.

Question 23AI Strategy and Governance

A governance team is prioritizing which AI systems to review first. Which prioritization logic is most defensible?

  1. Review systems alphabetically by project name
  2. Review highest-impact, highest-uncertainty systems first Correct answer.
  3. Review only systems built with deep learning
  4. Review the cheapest systems first to clear the backlog

Explanation

Prioritizing by impact and uncertainty directs limited review capacity toward the systems most likely to cause significant harm and least understood. Alphabetical, cost-based, or model-type filters ignore risk and can leave the most consequential systems unreviewed.

Question 24AI Strategy and Governance

Which control most directly supports the right of individuals to understand and contest automated decisions that affect them?

  1. A policy requiring model retraining every month
  2. A quarterly internal audit of cloud spending
  3. A process for providing decision explanations and accepting appeals Correct answer.
  4. A rule that all models must use open weights

Explanation

Contestability requires that affected individuals receive a meaningful explanation and have a channel to challenge the outcome. Spending audits, retraining cadence, and weight openness do not provide individuals with explanation or appeal mechanisms.

Question 25AI Strategy and Governance

An organization discovers that a generative AI assistant used for internal research can reproduce sensitive content from its training data. Which governance response is most appropriate first?

  1. Publish the finding in a press release before analysis
  2. Expand access to all employees to gather more feedback
  3. Restrict access and investigate the exposure while assessing scope Correct answer.
  4. Take no action because the model is only used internally

Explanation

Potential leakage of sensitive content is a containment event, so limiting access and investigating scope takes priority before any broader communication or expansion. Expanding access or waiting because use is internal increases exposure, and premature disclosure risks harm before facts are known.

Question 26AI Strategy and Governance

An organization's board asks for a one-page statement of how AI will create value, which risks it will accept, and which uses it will refuse. Which artifact most directly answers this request?

  1. A data retention schedule for training corpora
  2. A model card describing the architecture of the production model
  3. An incident response runbook for model outages
  4. An AI strategy that states objectives, risk appetite, and explicit exclusions Correct answer.

Explanation

An AI strategy is the governing statement of intent: it sets objectives, defines how much risk the organization will accept, and names the uses it will not pursue. Model cards, retention schedules, and runbooks are operational artifacts that implement decisions rather than declare them.

Question 27AI Strategy and Governance

A newly appointed AI governance lead must decide where the function should report so that it can constrain business units that outrank it. Which reporting line best supports that objective?

  1. Into the data science team that builds most of the models
  2. Into an enterprise risk or compliance function with board-level visibility Correct answer.
  3. Into the largest business unit that uses the most AI models
  4. Into the procurement group that negotiates vendor contracts

Explanation

Independence and escalation authority are what let a governance function challenge powerful business units. Reporting into enterprise risk or compliance, with a direct line to the board or a board committee, preserves that authority. Embedding governance inside a model-building or revenue-owning team creates a conflict of interest.

Question 28AI Strategy and Governance

Which pairing correctly matches a governance body with its primary decision right?

  1. AI review board: approves or rejects high-risk use cases before deployment Correct answer.
  2. Data engineering team: sets the enterprise risk appetite for AI
  3. Internal audit: authors the AI strategy and product roadmap
  4. Model owner: grants regulatory exemptions for automated decisions

Explanation

A review board exists to make gating decisions on use cases that carry elevated risk, approving, conditioning, or rejecting them before launch. Risk appetite belongs to executive leadership, strategy belongs to the business, and no internal team can grant regulatory exemptions.

Question 29AI Strategy and Governance

A firm wants every AI use case to have a named individual who answers for its outcomes, even though dozens of engineers contribute. Which assignment satisfies this requirement?

  1. A rotating on-call engineer drawn from the platform team
  2. A single accountable owner with authority over the use case Correct answer.
  3. The vendor whose foundation model powers the system
  4. The end users who interact with the system daily

Explanation

Accountability requires a named individual with the authority and resources to change or stop the system. Rotating staff, vendors, and end users may contribute or be affected but cannot be held accountable for the organization's decision to deploy.

Question 30AI Strategy and Governance

A regulator asks an organization to show that it applied proportionate scrutiny to a low-risk internal AI tool versus a high-risk hiring model. Which governance mechanism most directly produces that differentiated evidence?

  1. A blanket policy prohibiting all internal AI tools
  2. A monthly newsletter summarizing AI news for staff
  3. A tiered risk classification with review depth tied to each tier Correct answer.
  4. A single annual attestation signed by all model owners

Explanation

Proportionality is demonstrated by a classification scheme that assigns review depth, documentation, and approval level according to risk tier. One uniform attestation, blanket bans, and awareness communications do not show that scrutiny scaled with risk.

Question 31AI Strategy and Governance

A governance committee is drafting tolerances for an AI system that screens job applicants. Which formulation is a usable risk appetite statement rather than a vague aspiration?

  1. We will use AI responsibly and ethically in hiring
  2. Adverse impact ratio must stay at or above 0.80, verified each quarter Correct answer.
  3. Hiring managers should feel comfortable with the model's outputs
  4. The model should be as accurate as reasonably possible

Explanation

A usable risk appetite statement names a measurable threshold, the metric, and the verification cadence so that breaches trigger action. Aspirational language and subjective comfort tests give no basis for a pass or fail determination.

Question 32AI Strategy and Governance

Leadership wants AI investments tied to measurable business outcomes rather than experimentation for its own sake. Which portfolio practice best enforces that discipline?

  1. Approving initiatives only when the model uses the newest architecture
  2. Requiring each initiative to state a baseline metric, target, and review date before funding Correct answer.
  3. Funding every proposed proof of concept to encourage exploration
  4. Ranking initiatives solely by the size of the requesting department

Explanation

Tying funding to a baseline, a target, and a scheduled review makes value realization testable and allows stop-or-scale decisions. Funding everything, weighting by department size, or selecting on architecture fashion disconnects spend from outcomes.

Question 33AI Strategy and Governance

An organization must decide whether to build a bespoke model or adopt a vendor API for a customer service assistant. Which governance criterion is most decision-relevant alongside cost?

  1. The ability to audit, explain, and control changes to the underlying model Correct answer.
  2. The number of engineering blog posts published about each option
  3. The physical office location of the vendor's sales team
  4. The programming language used in the internal prototype

Explanation

Build-versus-buy decisions turn on how much visibility and control the organization retains over model behavior, change management, and explainability. Those capabilities determine whether the organization can meet its own governance and regulatory obligations.

Question 34AI Strategy and Governance

A company wants its AI strategy to remain credible as model capabilities shift. Which design choice best supports that durability?

  1. Publishing the strategy once and treating later revisions as a failure
  2. Delegating all strategy updates to the vendor account manager
  3. Stating durable principles and outcomes, with a scheduled mechanism to refresh tactics Correct answer.
  4. Committing to a fixed list of approved model vendors for five years

Explanation

Durable strategy separates stable intent, such as principles and target outcomes, from tactics that must adapt as technology and regulation change. A built-in refresh cadence keeps the document current without abandoning its core commitments.

Question 35AI Strategy and Governance

An audit finds that AI policy exists but engineers routinely bypass it because approval queues take weeks. Which governance response addresses the root cause most directly?

  1. Extending the approval queue to all internal tooling
  2. Assigning the policy to a new author for rewriting
  3. Adding a fast-track path for low-risk use cases with automated checks Correct answer.
  4. Issuing a stricter policy restating the same approval requirement

Explanation

When compliant behavior is slower than circumvention, people route around the control. Reducing friction for genuinely low-risk cases through automated, proportionate checks makes the compliant path the fastest path while preserving scrutiny where risk is high.

Question 36AI Strategy and Governance

Which statement best distinguishes an AI policy from an AI standard within a governance framework?

  1. A policy sets mandatory high-level requirements; a standard specifies how those requirements are met Correct answer.
  2. A policy applies to vendors; a standard applies to employees
  3. A policy is optional guidance; a standard is legally binding
  4. A policy describes model architecture; a standard describes training data

Explanation

In a typical hierarchy, policy states what must be true and why, while standards translate those mandates into specific, testable requirements and methods. Procedures then describe step-by-step execution. The other pairings misstate scope or audience.

Question 37AI Strategy and Governance

A multinational must reconcile a global AI principle of non-discrimination with local laws that differ in how they define protected groups. Which approach is most defensible?

  1. Suspend the principle in jurisdictions where the law is unclear
  2. Let each country office define its own principles independently
  3. Apply the least stringent local rule everywhere to reduce cost
  4. Set a global baseline that satisfies the strictest applicable requirement, then document local adjustments Correct answer.

Explanation

A common pattern is to adopt a global floor aligned to the most demanding applicable regime and then record justified local variations. This preserves consistent commitments while respecting genuine legal differences, and the documentation makes deviations reviewable.

Question 38AI Strategy and Governance

An organization wants assurance that its AI governance controls operate as designed, not merely that they are documented. Which activity provides that assurance?

  1. Testing a sample of deployed models against the stated controls Correct answer.
  2. Adding the controls to the employee onboarding deck
  3. Publishing the control catalog on the intranet
  4. Renaming the control owner to a senior vice president

Explanation

Assurance comes from testing actual operation against the documented design, for example sampling deployed systems and checking evidence of the control. Publication, training, and title changes affect awareness and authority but do not verify that controls function.

Question 39AI Strategy and Governance

A board member asks how AI governance should relate to the organization's existing enterprise risk management. Which position is most coherent?

  1. AI risks should be registered and treated inside the existing enterprise risk framework Correct answer.
  2. AI risk should be tracked only when a regulator requests it
  3. AI governance should operate entirely outside formal risk reporting
  4. AI governance should replace enterprise risk management for technology topics

Explanation

AI risk is a category of enterprise risk, so it should be recorded in the same register, scored with consistent methods, and reported through the same escalation channels. A parallel or episodic system fragments oversight and hides aggregate exposure from leadership.

Question 40AI Strategy and Governance

Which competency profile best equips an AI governance function to challenge both technical and business claims?

  1. Deep expertise in a single modeling framework with no policy background
  2. General administrative skills with no AI-specific training
  3. Strong sales experience with no exposure to data or law
  4. Blended literacy in AI methods, risk management, and applicable regulation Correct answer.

Explanation

Effective challenge requires enough technical literacy to interrogate model claims, enough risk expertise to judge materiality, and enough legal knowledge to identify obligations. A single-discipline profile leaves one dimension of the argument unchecked.

Question 41AI Business Case and Value Realization

A team wants to justify an AI investment by comparing the projected benefits against the total cost of ownership over three years. Which financial metric most directly expresses that comparison as a single decision-ready number?

  1. Payback period measured in months
  2. Total cost of ownership expressed in absolute currency
  3. Gross revenue attributable to the AI system
  4. Net present value of the discounted cash flows Correct answer.

Explanation

Net present value discounts future benefits and costs to today's value and nets them into one figure, which is the standard basis for go/no-go investment decisions. Absolute cost, payback period, and gross revenue each capture only part of the picture and cannot alone show whether benefits exceed costs.

Question 42AI Business Case and Value Realization

An organization classifies an AI initiative as a 'quick win' in its portfolio. Which characteristic best defines a quick win in AI value realization terms?

  1. High technical complexity that builds reusable platform capability
  2. Low effort and fast time-to-value with modest but demonstrable benefit Correct answer.
  3. A research project exploring unproven model architectures
  4. High strategic impact delivered over a multi-year roadmap

Explanation

A quick win is defined by the ratio of speed and low effort to visible benefit, which builds momentum and stakeholder confidence. Multi-year high-impact work, complex platform builds, and pure research are the opposite of fast, low-effort delivery.

Question 43AI Business Case and Value Realization

A retail bank estimates that an AI fraud-detection model will reduce losses by $4 million per year. Before including this in the business case, what should the analyst confirm about the figure?

  1. That the vendor has published a similar case study
  2. That the model uses a state-of-the-art architecture
  3. That the benefit is baselined against current loss rates and is attributable to the AI system Correct answer.
  4. That the finance team has approved the model's hyperparameters

Explanation

Benefits in a business case must be baselined against the current state and credibly attributed to the intervention, otherwise the value estimate is inflated or unverifiable. Architecture choice, vendor marketing, and hyperparameter approval do not establish that the $4 million is real and caused by the AI.

Question 44AI Business Case and Value Realization

Which cost category is most often underestimated when teams build the total cost of ownership for a production AI system?

  1. The one-time cost of the cloud subscription sign-up
  2. Initial model training compute
  3. The purchase price of developer laptops
  4. Ongoing data labeling, monitoring, and model maintenance after deployment Correct answer.

Explanation

Post-deployment costs such as data labeling, drift monitoring, retraining, and support typically recur for the life of the system and are frequently omitted from early estimates. One-time training compute is usually visible up front, while sign-up fees and laptops are trivial or irrelevant.

Question 45AI Business Case and Value Realization

A business case claims that an AI assistant will cut average handling time by 30 percent. Which evidence would most strengthen this claim for an approver?

  1. An estimate produced by the project sponsor without documentation
  2. A comparison with a competitor's press release about their AI program
  3. A pilot result showing the measured reduction on a representative sample of live calls Correct answer.
  4. A statement that the vendor's model benchmarks well on public leaderboards

Explanation

A controlled pilot on representative production-like data provides direct empirical support for the claimed operational improvement. Model benchmarks measure general capability rather than this workflow, undocumented sponsor estimates lack rigor, and competitor press releases say nothing about this organization's processes.

Question 46AI Business Case and Value Realization

An AI program has several candidate use cases. Leadership wants to rank them using a value-versus-feasibility framework. Which pairing of dimensions best matches that framework?

  1. Model accuracy versus training dataset size
  2. Cloud spend versus headcount
  3. Expected business value versus organizational and technical feasibility Correct answer.
  4. Number of stakeholders versus number of data sources

Explanation

A value-versus-feasibility matrix plots initiatives by the benefit they can create against how achievable they are given data, skills, and process readiness. Accuracy, dataset size, stakeholder counts, and spend are inputs or attributes but do not form the classic prioritization axes.

Question 47AI Business Case and Value Realization

A team reports that its AI model achieves 95 percent accuracy, and the sponsor treats this as proof of business value. What is the most important gap in this reasoning?

  1. Model accuracy does not by itself establish that the metric aligns with business outcomes or that the improvement is actionable Correct answer.
  2. Accuracy should always be reported as a percentage with two decimals
  3. The team should have reported the model's parameter count instead
  4. Accuracy is only meaningful for regression models

Explanation

A technical metric like accuracy becomes a value claim only when it is tied to a business outcome and to decisions or actions that change as a result. High accuracy on an unrepresentative or irrelevant metric can coexist with little or no business benefit.

Question 48AI Business Case and Value Realization

Which approach best converts an AI efficiency gain into a realized financial benefit in a business case?

  1. Assume the saved hours automatically become cash savings
  2. Report the efficiency gain as revenue in the first year
  3. Plan how freed capacity will be redeployed or how headcount and cost lines will actually change Correct answer.
  4. Defer all benefit recognition until the model is retired

Explanation

Efficiency gains only become financial value when there is an explicit plan for redeploying capacity or reducing cost, otherwise the saving is theoretical. Automatically booking saved hours as cash or revenue overstates benefit, and deferring recognition discards the value entirely.

Question 49AI Business Case and Value Realization

An organization is deciding whether to build an AI capability in-house or buy a vendor solution. Which factor most directly favors buying in the short term?

  1. The capability is a core differentiator the organization intends to protect
  2. The organization wants full control over model internals and data handling
  3. The organization has a large ML team seeking challenging work
  4. Time-to-value is critical and the required capability is available as a mature product Correct answer.

Explanation

Buying accelerates time-to-value when a mature product already meets the need, since it avoids build lead time and staffing. Control needs, differentiation strategy, and internal team utilization are reasons that typically favor building or customizing.

Question 50AI Business Case and Value Realization

A business case for an AI customer-service tool lists reduced response time as the primary benefit. Which additional benefit is most defensible and commonly cited alongside it?

  1. Guaranteed increase in market share
  2. Higher customer satisfaction from faster, more consistent resolution Correct answer.
  3. Automatic compliance with every regulation
  4. Elimination of all customer complaints

Explanation

Faster and more consistent responses plausibly improve customer satisfaction, which is a measurable and commonly tracked benefit. Market-share gains, elimination of complaints, and automatic regulatory compliance are overclaims that cannot be guaranteed by a response-time improvement.