51 practice questions for Domain 2 of the AWS Certified AI Business Strategist (AIB-C01) exam, which makes up 28% of its scored content. Your answers count towards one score and one timer for the whole exam.
Domain 2: AI Strategy and Business Value Creation
46. A company must build a business case for an AI customer-service assistant. Which set of measures best demonstrates business value?
Answer and explanation
Answer: B. A defensible business case establishes a pre-deployment baseline, defines outcome KPIs that map to business results, and accounts for the full cost of ownership rather than inference alone. Model size and dataset size are technical attributes with no direct business meaning. Counting AWS services measures architectural complexity, not value. Headcount is an input cost, not a benefit measure.
47. An organization is deciding whether to build a custom model, buy a SaaS product, or use a foundation model through Amazon Bedrock. Which consideration most strongly favours the foundation model route?
Answer and explanation
Answer: A. When the model is a commodity and differentiation lies in proprietary data and workflow, using a managed foundation model with retrieval or fine-tuning delivers value fastest without rebuilding commodity capability. A genuinely unique algorithmic advantage argues for building. A use case with no data is not an AI use case. A regulatory prohibition on third-party models rules the option out entirely.
48. An AI pilot succeeded in one department, and leadership wants enterprise-wide rollout within a quarter. Which approach best manages risk while scaling?
Answer and explanation
Answer: C. Iterative scaling validates that value transfers to new contexts and lets governance, platform, and skills mature alongside adoption, which is how pilots become durable capability. Simultaneous enterprise deployment multiplies unvalidated assumptions across the organization. Confining a successful pilot forfeits the return. Rebuilding per department discards reuse and multiplies cost and inconsistency.
49. A company must prioritize among twelve proposed AI use cases. Which pair of dimensions is most appropriate for an initial screen?
Answer and explanation
Answer: D. Value against feasibility is the standard screen because it surfaces high-return cases the organization can actually deliver with the data and capability it has. Model size and training duration are implementation details. Service count and novelty measure complexity rather than worth. Sponsor enthusiasm is not evidence of value.
50. A pilot reduced average handling time by 18 percent for 200 agents. What is required to convert this into a credible ROI figure?
Answer and explanation
Answer: C. ROI requires monetising the benefit and subtracting the full cost of achieving it, including ongoing inference, integration, and the human cost of adoption. A percentage improvement without a monetary value is not a return. Invocation counts and benchmark accuracy are inputs to cost and quality rather than a return calculation.
51. An organization is deciding whether to build a custom model or use a managed foundation model. Which consideration most favours building?
Answer and explanation
Answer: D. Building is justified when the capability is itself the differentiator and cannot be bought. Speed to production, a commoditised use case, and limited expertise all argue for using a managed model instead.
52. Which cost element is most frequently underestimated in generative AI business cases?
Answer and explanation
Answer: C. Pilots run at low volume, so per-token inference cost at full production scale, together with integration into existing systems and the human effort of adoption, routinely dwarfs the pilot budget. Model selection is a small one-off effort. Account and notebook costs are trivial by comparison.
53. A strategist must define success measures for an AI-assisted document review tool before the pilot begins. Which set is most appropriate?
Answer and explanation
Answer: D. A baseline makes improvement measurable, outcome targets tie the tool to the business result, and adoption reveals whether the benefit is actually being realised. Document counts measure volume rather than value. Feature counts are a technical detail. Uptime is necessary but does not indicate business benefit.
54. Which approach best manages a portfolio of AI initiatives across a large organization?
Answer and explanation
Answer: C. Stage gates concentrate investment on initiatives that prove value and stop those that do not, which is how portfolio risk is managed. Equal funding spreads resource thinly regardless of merit. Backing one opportunity concentrates risk. Deferring everything forgoes learning and momentum.
55. A company's proprietary data is its main advantage in a use case where the underlying model is commodity. What does this imply for strategy?
Answer and explanation
Answer: C. Where the model is commodity and the data is the differentiator, the returns come from making that data usable and well governed while renting the model. Pre-training a proprietary model spends heavily on the commodity part. Avoiding AI forgoes the advantage. Licensing the data away removes the differentiator.
56. A pilot succeeded but the business case for scaling is weak because the benefit accrues to a small team. What is the appropriate response?
Answer and explanation
Answer: C. A capability that works but serves too few people is often worth extending to adjacent processes where the same investment earns more, which is how the case becomes viable. Scaling on a weak case wastes capital. Abandoning discards proven capability. Degrading quality to fit the numbers undermines the benefit that justified it.
57. Which statement about time to value is most accurate when comparing prompt engineering, retrieval augmentation, and fine-tuning?
Answer and explanation
Answer: A. The techniques form an escalating ladder of cost and effort, and starting at the cheapest rung often removes the need for the others. Fine-tuning involves data curation and a training run and is not fastest. Retrieval is usually less effort than fine-tuning. The three differ substantially in investment.
58. An executive asks why the company should not simply wait until AI capabilities mature further. Which argument is strongest?
Answer and explanation
Answer: A. The durable constraint is organizational rather than technological, and the prerequisites compound slowly, so starting now shortens the path whenever the technology settles. Model prices have fallen substantially and are likely to continue. Competitive gaps can be closed. Waiting has a real opportunity cost.
59. A vendor proposes an AI product priced per seat. Which question most directly tests whether the pricing fits the company's usage?
Answer and explanation
Answer: D. Per-seat pricing decouples cost from usage, which can favour or penalise the buyer depending on intensity, so modelling cost at expected volume is the decisive test. Customer count, underlying model, and vendor age inform diligence but do not determine whether the pricing structure fits.
60. A strategist is mapping AI opportunities across the business. Which starting point is most productive?
Answer and explanation
Answer: C. Opportunity mapping starts from where work is repetitive and data already exists, because those are the places value is achievable rather than merely imaginable. Technology-first surveys produce solutions in search of problems. Wish lists collect enthusiasm rather than feasibility. Political visibility does not correlate with value or feasibility.
61. A pilot must be designed to produce a decision about whether to scale. Which design is most useful?
Answer and explanation
Answer: A. A pilot exists to produce a decision, so it needs a scope, a baseline, a threshold agreed in advance, and a date when the decision is made. Open-ended exploration never resolves. Covering every department is a rollout rather than a pilot. Judging by impression invites confirmation bias.
62. Which cost components should a total cost of ownership estimate for a generative AI application include?
Answer and explanation
Answer: A. TCO must include the whole delivery and operating cost, of which inference is often a minority. Inference alone understates the figure substantially. Account and support costs are minor. A vendor list price omits everything the business must do to realise value.
63. Two proposals compete for funding: one automates a task saving 3 percent of a large cost base, the other enables a new revenue line with uncertain demand. How should a strategist frame the choice?
Answer and explanation
Answer: C. The two have different risk and return profiles, so the decision is about portfolio balance rather than a single comparison. Headline figures ignore probability. Always choosing low risk forgoes growth. Technical interest is not a value criterion.
64. A vendor's AI product is priced per outcome, such as per resolved support ticket. What should a buyer verify?
Answer and explanation
Answer: A. Outcome-based pricing hinges entirely on the definition and measurement of the outcome, so the definition and dispute process determine what is actually paid. Unit price is meaningless without the definition. The underlying model and vendor size do not determine billing.
65. Which measure best captures the productivity effect of an AI assistant on knowledge workers?
Answer and explanation
Answer: D. Productivity is a change in how long work takes and how good it is, measured against a baseline for comparable work. Satisfaction and usage volume indicate engagement rather than output. Licence count is an input cost.
66. A company's data is fragmented across systems with inconsistent definitions. What does this imply for its AI strategy?
Answer and explanation
Answer: D. Most AI value depends on trustworthy, joinable data, so fragmentation is a gating constraint rather than something AI resolves. Adopting AI on inconsistent data produces unreliable results. Both predictive and generative use cases are affected. External data does not fix internal inconsistency.
67. Which situation most justifies fine-tuning rather than retrieval augmentation from a business perspective?
Answer and explanation
Answer: A. Fine-tuning encodes durable behaviour into the model, whereas retrieval supplies changing facts at query time. Daily-changing documents argue for retrieval. Citation requires retrieved sources. Fine-tuning does not inherently reduce inference cost and adds training cost.
68. An initiative shows a strong return in the pilot but the benefit depends on staff changing an established workflow. What should the business case include?
Answer and explanation
Answer: C. When benefit depends on behaviour change, the case must cost that change and reflect the risk of partial adoption. Technology cost alone understates the investment. Assuming immediate full adoption is unrealistic. A mandate is a tactic rather than a costed plan and often produces compliance without benefit.
69. Which consideration should weigh most heavily when choosing between two foundation model providers for a long-term platform?
Answer and explanation
Answer: B. Model capability changes quickly, so the durable decision is architectural: keeping the application loosely coupled to any one provider preserves the ability to switch. Recency, size, and trial length are transient factors that will not hold over a platform's life.
70. A department proposes an AI project whose benefit accrues mostly to a different department. What governance mechanism addresses this?
Answer and explanation
Answer: C. Misaligned cost and benefit is a common reason valuable cross-functional projects stall, and an agreed attribution model removes the disincentive. Making the proposer bear all cost discourages exactly the projects worth doing. Rejecting them forgoes value. Equal allocation ignores where benefit actually lands.
71. Which approach best sustains competitive advantage from AI when competitors have access to the same models?
Answer and explanation
Answer: B. When models are commodity, advantage comes from assets competitors cannot copy quickly: proprietary data, deep workflow integration, and the learning that accrues from operating a system at scale. Model size is purchasable. Announcement order is not an advantage. Architectural secrecy is hard to maintain and rarely decisive.
72. An AI initiative's costs are rising faster than its usage. What is the most likely cause to investigate first?
Answer and explanation
Answer: A. Generative AI cost scales with tokens rather than user count, so cost outpacing usage usually means each interaction has grown more expensive. Headcount affects usage volume rather than cost per interaction. Region and storage are minor contributors to a generative workload's cost profile.
73. An organization plans to measure the effect of an AI initiative on handling time. What must be captured first?
Answer and explanation
Answer: D. Without a baseline there is nothing to measure improvement against, so the baseline must be captured before deployment. A vendor benchmark describes their environment. Budget and user count are inputs rather than an outcome measure.
74. Which pair of benefits should an AI business case include? (Select TWO.)
Answer and explanation
Answer: B, C. A complete business case captures both quantifiable financial effects and the intangible benefits that often carry most of the value. Model count, Region, and language are implementation details with no place in a business case.
75. Which signal is a leading indicator that an AI initiative is likely to succeed?
Answer and explanation
Answer: B. Adoption by intended users predicts eventual value, since an unused system delivers none. Model count, budget size, and vendor count measure activity and spend rather than likely outcome.
76. Which use of AI is most likely to create a durable competitive advantage?
Answer and explanation
Answer: B. Advantage is durable when it rests on something competitors cannot easily copy, which proprietary data and processes provide. A widely available tool is available to everyone. Model size is not a differentiator. An announcement is not a capability.
77. An industry is at an early stage of AI maturity with few established competitors using AI. Which investment posture is appropriate?
Answer and explanation
Answer: D. Early maturity means high uncertainty, so investment should build capability and learning without over-committing. Matching an unrelated industry's spend ignores context. Deferring entirely forfeits the learning. Committing the whole budget bets everything on an unproven direction.
78. How can AI transform a business model rather than only improving an existing process?
Answer and explanation
Answer: C. Business model transformation means doing something newly possible, such as a service that was previously uneconomic. Cost, speed, and accuracy improvements are valuable but they optimise the existing model rather than changing it.
79. An AI strategy must be aligned with business objectives. Which approach achieves this?
Answer and explanation
Answer: D. Alignment begins with business outcomes and works toward capability. Starting from available technology, competitor announcements, or team preference produces solutions looking for problems.
80. A portfolio of candidate AI use cases must be prioritized. Which criteria should be applied?
Answer and explanation
Answer: A. Value, feasibility, and risk together sequence a portfolio rationally. Proposal order, sponsor seniority, and novelty are not decision criteria.
81. An organization must decide whether to build an AI capability internally or buy a vendor solution. Which consideration favours building?
Answer and explanation
Answer: C. Differentiation resting on proprietary data justifies building. A commodity function, absent internal capability, and a short timeline all favour buying.
82. A business process is being considered for AI-enabled transformation rather than incremental improvement. Which indication supports transformation?
Answer and explanation
Answer: C. Transformation means the process is reconceived because something previously impossible is now viable. Speed, error rate, and cost point to incremental improvement of the existing process.
83. An AI initiative's return on investment must be calculated. Which costs must be included?
Answer and explanation
Answer: C. A defensible return calculation includes the full lifecycle cost including ongoing operation and governance. Any narrower definition understates the investment.
84. Which measure indicates that an AI initiative has improved productivity?
Answer and explanation
Answer: C. Productivity is measured against a pre-deployment baseline for the affected task. Model counts, data volume, and training numbers describe activity rather than productivity.
85. An executive asks how quickly an AI initiative will pay back its investment. Which information is required to answer?
Answer and explanation
Answer: D. Payback requires investment, periodic benefit, and the crossover point. Token cost, user count, and model accuracy are inputs that inform but do not answer the question.
86. Which factor most commonly causes an AI initiative to deliver less value than projected?
Answer and explanation
Answer: C. Adoption is the most common gap between projected and realised value. Parameter count, compute capacity, and model count are technical inputs rather than the usual cause.
87. An organization's proprietary data is being considered as a basis for AI advantage. Which characteristic makes it genuinely advantageous?
Answer and explanation
Answer: B. Advantage rests on data competitors cannot replicate. Format, volume, and recency improve usability without conferring exclusivity.
88. An industry is mature in AI adoption with most competitors already deploying comparable capabilities. Which posture is appropriate?
Answer and explanation
Answer: D. In a mature market some capability is table stakes and differentiation must come from genuine advantage. Avoiding investment cedes the market, matching everything wastes resources, and waiting compounds the gap.
89. An AI capability that competitors could replicate within months is being proposed as a strategic differentiator. Which assessment is appropriate?
Answer and explanation
Answer: B. A replicable capability is a temporary advantage requiring a follow-on plan. First-to-market alone is not durable, abandoning forgoes the interim benefit, and delaying forgoes it entirely.
90. Which measure indicates that an AI initiative has reduced operational cost?
Answer and explanation
Answer: C. Cost per transaction against a pre-deployment baseline measures the effect. Initiative spend, process count, and training numbers describe inputs.
91. An AI initiative's benefits are largely intangible. How should they be presented in a business case?
Answer and explanation
Answer: D. Using a defensible proxy where available and naming the remainder honestly is credible. Excluding them understates value, equating them overstates it, and a fixed proportion is arbitrary.
92. Which practice establishes whether an AI initiative caused an observed improvement?
Answer and explanation
Answer: D. A control or baseline comparison isolates the initiative's effect. Post-deployment observation, user impressions, and industry benchmarks all leave other causes unexcluded.
93. Which indicator suggests an AI initiative will not realise its projected value?
Answer and explanation
Answer: A. Low adoption means the value is not realised regardless of technical quality. Accuracy short of theoretical maximum, timeline, and infrastructure cost affect but do not determine realisation.
94. An organization's AI capability is built entirely on a third-party vendor's platform. Which strategic risk does this create?
Answer and explanation
Answer: D. Vendor dependence constrains differentiation and negotiating position. Accuracy, cost, and latency are operational rather than strategic risks.
95. Which approach allows an organization to differentiate using AI when competitors use the same foundation models?
Answer and explanation
Answer: A. Proprietary data and processes are what competitors cannot copy. Model size, timing, and breadth of deployment are all replicable.
96. An organization must assess whether an AI capability is worth building given its industry's competitive position. Which analysis is appropriate?
Answer and explanation
Answer: A. Customer expectation, competitive position, and distinctive capability together determine whether to build. Technology maturity, announcements, and budget are inputs rather than the analysis.