AI Strategy and Business Value: 5 AIB-C01 Questions with Full Explanations
AWS AI Business Strategist (AIB-C01) · AI Strategy and Business Value · 5 questions with answers and explanations
Here are 5 AIB-C01 practice questions on AI Strategy and Business Value. It carries about 28% of the exam score. Each answer explains the correct choice and what is wrong with the others.
🧭 Before you start
This domain puts you in a leadership seat. A list of candidate projects, a stated business goal, a rising bill or a partner dependency sets the scene, and you choose how to prioritize, measure or protect the investment. Strong answers tie AI work to a measurable outcome; weak ones reward novelty, visibility or the sheer number of projects.
Watch for these while you work:
• Rank candidate use cases on two axes at once, the value they bring and how feasible they are, and be wary of options that favor the most ambitious idea.
• Link each initiative to the metric the goal names. A useful project that moves a different number does not support that goal.
• Check cost and value claims for what is left out, such as data preparation, staffing or change management, and judge progress against the original business case.
Question 1 of 5
An insurer has 25 proposed GenAI use cases and budget for three in the next year. Which prioritization approach is MOST appropriate?
Show answer
✅ Answer: D
Difficulty: advanced
Prioritizing a large AI backlog
With 25 ideas and budget for three, the question is which projects will deliver real value and can actually be delivered. A common method is a value-feasibility matrix:
• Business value: revenue impact, cost savings, risk reduction, customer experience, strategic fit.
• Feasibility: data availability and quality, technical complexity, integration effort, skills, regulatory and responsible AI risk, time to value.
Each use case is scored on both axes, for example 1 to 5, and plotted on a 2x2 grid.
• High value, high feasibility: start here; these become early wins that fund and justify the program.
• High value, low feasibility: plan for later, after data or platform groundwork.
• Low value: deprioritize, however interesting the technology.
Example
An internal claims-document summarizer with good data and moderate risk might score high on both axes, while a fully automated customer-facing claims decision engine may be valuable but low on feasibility because of regulatory risk.
Why each option works or fails
• ✗ A. Most advanced technology: technical novelty is not business value, and cutting-edge approaches are often the least feasible to deliver in a year.
• ✗ B. Every department in parallel: no central criteria means more than three projects compete for a three-project budget, spreading funds thin with no way to compare value.
• ✗ C. Highest-risk customer-facing first: starting with the riskiest use case raises the chance of a visible failure that can stall the whole program.
• ✓ D. Score on value and feasibility (correct): consistent criteria let the insurer pick the three use cases most likely to deliver measurable value within budget and build momentum for the rest.
🎯 Key idea: Prioritize AI use cases on value times feasibility and start with those high on both, not on novelty or visibility.
Question 2 of 5
A utility company's leadership wants AI to support its goal of cutting unplanned outages by 20% in two years. Which initiative aligns most directly with that goal?
Show answer
✅ Answer: A
Difficulty: applied · Topic: Amazon SageMaker AI
Aligning AI with a business goal
Strategy alignment means starting from the goal, here cut unplanned outages by 20% in two years, and asking which AI use case moves that number directly.
Predictive maintenance
Predictive maintenance uses sensor readings, inspection records and failure history to estimate which assets are likely to fail soon.
• A model scores transformers, lines or pumps by failure risk.
• Crews repair or replace the highest-risk equipment during planned work, before it fails.
• Success is measured with the same metric leadership cares about: the count of unplanned outages.
Example: vibration and temperature data flag a transformer running hot; it is replaced during a scheduled window instead of failing during a storm.
Good ideas, wrong goal
Customer service or marketing tools can be worthwhile, but they belong to other goals. Funding them under the outage objective would not move the target metric.
Why each option works or fails
• ✓ A. Predictive maintenance (correct): flagging equipment that is likely to fail lets crews fix it in advance, which directly reduces unplanned outages, the exact metric in the stated goal.
• ✗ B. Billing chatbot: shorter call-center wait times may improve customer experience, but answering billing questions does nothing to keep equipment running.
• ✗ C. Generated newsletter: automating a customer newsletter saves writing effort but has no link to grid reliability or outage counts.
• ✗ D. Marketing image generator: visuals for advertising campaigns support marketing goals, not the physical reliability of the network.
🎯 Key idea: Pick the AI initiative whose output directly changes the goal's metric; worthwhile projects aimed at other goals do not count toward this one.
Question 3 of 5
A company wants to track whether its AI investments are paying off across the portfolio. Which practice helps most?
Show answer
✅ Answer: B
Difficulty: applied
Measuring AI return across a portfolio
Each AI initiative should start with a business case: expected benefits (time saved, revenue, error reduction), costs and a timeline. Tracking pays off only when actual results are compared with that plan.
A portfolio review in practice
• Every quarter, list each initiative with its planned and measured value, cost to date and status.
• Example: a claims assistant planned to save 10,000 hours this year has saved 4,000 by mid-year, on track; a chatbot planned to cut calls by 15% has cut 2%, so it needs a fix or should be stopped.
• Use the review to scale winners, fix laggards and stop initiatives that will not pay off.
Activity counts or spending totals alone cannot show this.
Why each option works or fails
• ✗ A. Count initiatives launched: the number of projects measures activity, not value. Many launches can coexist with poor returns.
• ✓ B. Regular review against each business case (correct): comparing measured results with planned value initiative by initiative shows which investments pay off and where to reallocate money.
• ✗ C. Employee sentiment survey: feelings about AI adoption can inform change management, but they do not measure financial return on specific investments.
• ✗ D. Yearly spending total without links: knowing total spend without tying it to outcomes shows cost only, so return on investment cannot be calculated.
🎯 Key idea: Track AI investments by regularly comparing each initiative's measured results with its business case, not by counting projects or totaling spend.
Question 4 of 5
A company builds an AI feature on top of a partner's platform. Which TWO risks should the strategy address? (Select TWO)
Show answer
✅ Answer: A, E
Difficulty: advanced
Risks of building on a partner's platform
Building an AI feature on another company's platform, such as a model provider or SaaS tool, speeds delivery but creates dependency.
• Roadmap risk: the partner may change or retire features or models your product relies on.
• Pricing risk: price changes affect your margins.
• Availability risk: their outages become your outages.
• Portability (lock-in) risk: if the partnership ends, can you export your data, prompts, fine-tuning data and configurations and move to another provider?
How strategies address this
Contracts with service levels and notice periods, abstraction layers so models can be swapped, keeping copies of your data and prompt libraries, and a documented exit plan. For example, routing calls through a service that offers several models makes switching models easier.
Why each option works or fails
• ✓ A. Dependence on roadmap, pricing and availability (correct): changes or outages on the partner's side directly affect the company's feature, cost and reliability.
• ✗ B. Executives attending the conference: event attendance has no effect on the technical or commercial risk of the dependency.
• ✗ C. Brand colors matching: visual branding is cosmetic and unrelated to strategic platform risk.
• ✗ D. Same time zone: time zones rarely matter for platform services, which are consumed through APIs and support agreements; this is not a strategic risk.
• ✓ E. Portability of data and prompts (correct): the ability to move data, prompts and configurations limits lock-in and protects the company if the partnership ends.
🎯 Key idea: When building on a partner, manage dependency (roadmap, pricing, availability) and portability (data and prompts), not cosmetic or social factors.
Question 5 of 5
A team's chatbot bill doubled after it began pasting full policy manuals into every prompt, and long replies are sometimes cut off mid-sentence. Which concept explains both effects?
Show answer
✅ Answer: B
Difficulty: advanced · Topic: Amazon Bedrock
What a token is
LLMs do not read characters or whole words; they read tokens, chunks of text such as a word, part of a word or punctuation. As a rough rule for English, one token is about three-quarters of a word, so 1,000 words is roughly 1,300 tokens (the exact number depends on the model's tokenizer).
Why tokens drive cost and limits
• Pricing: on-demand inference in Amazon Bedrock is billed per input token and per output token.
• Context window: each model has a maximum number of tokens it can consider at once (prompt plus response).
• Max output tokens: the request sets a cap such as maxTokens; when the reply hits it, generation stops, even mid-sentence, and the API reports a stop reason like max_tokens.
Example: pasting a 40-page manual (about 25,000 tokens) into every prompt multiplies the input tokens billed per request, which explains the doubled bill.
Why each option works or fails
• ✗ A. Parameters: parameter count describes model size and influences the per-token price of a model, but you are not billed per parameter used, and parameters do not cause truncated replies.
• ✓ B. Tokens (correct): input tokens from the pasted manuals raised the bill, and the output token limit caused replies to stop mid-sentence. Both effects come from the same unit that billing and limits are measured in.
• ✗ C. Embeddings: embeddings are vectors produced by separate embedding models, typically for search and RAG. A chat request is not billed by vector size, and embeddings do not cut off generated text.
• ✗ D. Epochs: epochs are passes over training data during training or fine-tuning. Inference does not retrain the model on each prompt, so epochs play no role in per-request cost.
🎯 Key idea: Long prompts cost more and long answers get cut off because both billing and limits are counted in tokens, not in parameters or epochs.
📊 This set at a glance
Exam: AIB-C01 · 85 questions · 130 minutes · passing score 700
This domain: about 28% of the exam, the largest of its 4 domains
This set: 5 questions · 3 advanced, 2 applied · 4 scenario, 1 concept check · 1 with more than one answer
Services and topics: Amazon SageMaker AI, Amazon Bedrock
🔗 Related practice sets
• 5 Questions to Test Your AI Governance and Responsible AI Knowledge (AIB-C01)
• All AIB-C01 practice sets
• Every exam on this blog
📚 Beyond this set
More AIB-C01 questions: https://hyunhp.github.io/cert-bank/#exam=AIB-C01
Study guide: https://hyunhp.github.io/cert-bank/study.html#exam=AIB-C01
These are original practice questions, not real exam questions, and cert-bank is not affiliated with any certification vendor.
Original practice questions written for this blog, not exam dumps.
Comments
Post a Comment