# Free AIB-C01 practice questions

> 10 free AIB-C01 practice questions with the correct answer and an explanation.

Page: https://devopscodelab.com/exams/aws-aib-c01/free-practice-questions/
Practice exams: https://devopscodelab.com/exams/aws-aib-c01/

### Question 1 of 10

An airline is choosing how to automate responses to passenger compensation claims. The rules for eligibility are defined by regulation, but passengers describe their situations in free-text messages with attachments. Which solution design is MOST appropriate?

- A. Handle all claims manually, because combining AI interpretation with regulated eligibility rules is not permitted
- B. Use AI to extract the facts from messages and attachments, apply the eligibility rules in rule-based logic, and route borderline cases to staff
- C. Use a generative AI model to read each claim and decide eligibility on its own judgment without encoded rules, with staff spot-checking a monthly sample
- D. Use rule-based keyword matching on the free text and attachments to decide eligibility automatically for every claim

**Answer:** B

**Explanation:** Combining AI to interpret unstructured inputs with deterministic rules for regulated eligibility uses each approach where it fits, and human review handles ambiguity. Keyword matching cannot reliably interpret varied free text and attachments. Hybrid AI-plus-rules designs are common and permitted, so manual handling forgoes the benefit. Letting a generative model decide eligibility on its own judgment makes regulated outcomes unpredictable, and a monthly spot-check catches errors only after customers are affected.

### Question 2 of 10

A retail executive must choose between a traditional ML classification model and a generative AI model for routing 200,000 monthly customer emails into eight fixed categories. The labelled history is large, categories rarely change, and per-message cost must be minimal. Which recommendation is MOST appropriate?

- A. Use rule-based keyword matching maintained by the support team, because the categories rarely change
- B. Use the largest available generative AI model, because its language understanding gives the highest accuracy on emails
- C. Train a traditional ML classifier on the labelled history; it is cheaper per prediction and suits fixed categories
- D. Use a generative AI model with a few-shot prompt containing examples of each category for every email

**Answer:** C

**Explanation:** With abundant labelled data, stable categories, and a strict cost target, a trained classifier is typically far cheaper per prediction while delivering strong accuracy. A few-shot generative prompt works but pays per-token inference costs on 200,000 emails a month without using the labelled history. The largest model adds even more cost without clear benefit for fixed classification, and keyword rules are brittle with varied email wording and are usually less accurate than a trained classifier.

### Question 3 of 10

A law firm pilots a generative AI tool to summarise contracts. Short contracts summarise well, but for 300-page agreements the summaries ignore clauses near the end. What is the MOST likely cause the program manager should investigate?

- A. The model's training data cutoff predates the agreements, so it cannot recognise newer clause types
- B. The tool's guardrails are filtering indemnity and termination language as sensitive content
- C. The contracts exceed the model's context window, so later content is truncated or poorly attended to
- D. The model needs fine-tuning on legal vocabulary before it can interpret clauses in long agreements

**Answer:** C

**Explanation:** Models can process only a limited number of tokens at once, so very long documents may be truncated or summarised unevenly; splitting the document or choosing a model with a larger context window addresses this. Fine-tuning on vocabulary would not lift the length limit, and short contracts already summarise well. Guardrails filter harmful content rather than end-of-document clauses, and a training cutoff does not prevent the model from reading a document supplied in the prompt.

### Question 4 of 10

A regional grocery chain notices that two national competitors have launched AI-driven personalised promotions and are gaining loyalty-programme members. The chain's leadership wants to understand its position before investing. What should the strategy team do FIRST?

- A. Commission a technical audit of the chain's data warehouse to confirm it can support machine learning workloads
- B. Assess where AI gives rivals an advantage and where the chain's own data and customer relationships could differentiate it
- C. License the same personalisation vendor the competitors use so the chain can match their promotions within one quarter
- D. Run a quick pilot of AI-generated coupons in a few stores and decide on wider investment from the redemption rate

**Answer:** B

**Explanation:** Leadership wants to understand its position first, so a competitive assessment that shows where rivals gain from AI and where the chain's local knowledge and loyalty data could differentiate it should come before investment. Licensing the same vendor yields parity at best and is an investment decision made before analysis. A coupon pilot tests a tactic before the strategy is clear. A data-warehouse audit checks technical readiness but not competitive position.

### Question 5 of 10

Several AI vendors offer the same generative AI writing tool to every company in an industry. A publisher asks how AI can create a sustainable competitive advantage rather than temporary parity. Which answer is MOST accurate?

- A. Combining AI with proprietary data, unique workflows, and capabilities that competitors cannot easily replicate
- B. Accepting that no advantage is possible because the models are shared, and competing on price and distribution instead
- C. Adopting the generic tool before rivals do, so the publisher captures market share during the lead period before others catch up
- D. Standardising on the largest and most expensive frontier model, since higher capability per request keeps output ahead of rivals

**Answer:** A

**Explanation:** Generic tools available to everyone deliver parity; sustainable advantage comes from proprietary data, distinctive processes, and skills built around AI, which are hard to copy. Being first with a generic tool gives a short lead that disappears as rivals adopt it. Paying for the most expensive model is something any competitor can also buy, so it is not defensible. Advantage is still possible through what the company uniquely combines with shared models.

### Question 6 of 10

A bank plans to move its customer-service chatbot from one generative AI provider to another to reduce costs. Which consideration is MOST important to evaluate before the transition?

- A. Whether the new provider offers a larger context window and more languages than the current one
- B. The per-token price difference between providers, since cost reduction is the stated goal of the move
- C. Continuity and quality: test the new model on real customer questions and plan a phased cutover with rollback
- D. Whether the existing chatbot's code repository and prompt templates can be archived after the switch

**Answer:** C

**Explanation:** Moving between AI platforms can change answer quality, latency, and behaviour, so evaluating the new model on real questions and planning a phased cutover with rollback protects customers and continuity. A larger context window or more languages are features that may not matter for this workload. Comparing only token prices ignores migration effort, quality, and prompt rework, and archiving the old repository is housekeeping rather than a key risk consideration.

### Question 7 of 10

A European e-commerce company plans an AI system that screens job applicants. Under the EU AI Act's risk-based approach, how is this use MOST likely classified?

- A. Unacceptable risk, because automated profiling of applicants is among the practices the Act prohibits outright
- B. Minimal risk, because the employer makes the final hiring decision, so only voluntary codes of conduct apply
- C. Limited risk, requiring only that candidates are told they are interacting with an AI system or AI-generated content
- D. High-risk, which brings obligations such as risk management, data governance, human oversight, and documentation

**Answer:** D

**Explanation:** The EU AI Act lists AI used in employment, such as recruitment and candidate screening, among high-risk uses, which carry requirements including risk management, data quality, human oversight, and technical documentation. Prohibited practices are a narrower set, such as social scoring or emotion recognition in the workplace, and do not include candidate screening in general. Limited-risk transparency duties and minimal-risk voluntary codes apply to lower-impact uses, and a human making the final decision does not take recruitment screening out of the high-risk category.

### Question 8 of 10

An online retailer wants to use detailed browsing histories to personalise prices for individual shoppers, which marketing estimates would raise margin by 3 percent. Customer research suggests shoppers would view this as unfair if they knew. How should the business leader navigate this tradeoff?

- A. Drop all use of browsing data, including product recommendations, so no personalisation can be perceived as unfair
- B. Proceed with personalised pricing but leave it out of customer communications, since the privacy notice already covers browsing data
- C. Proceed, because a 3 percent margin gain outweighs the reaction, and run a limited A/B test to measure any churn
- D. Weigh the fairness, transparency, and trust risks against the gain with legal and ethics input, and consider less intrusive options

**Answer:** D

**Explanation:** When business goals conflict with responsible AI principles, leaders should weigh the benefits against fairness, transparency, legal, and trust risks with the right stakeholders and look for alternatives that keep most of the value. Proceeding and only measuring churn ignores real harm and legal exposure. Relying on a generic privacy notice while not disclosing personalised pricing violates transparency, and dropping all personalisation discards legitimate, low-risk value such as recommendations.

### Question 9 of 10

An accounts payable team currently keys invoice data by hand. After an AI extraction tool is introduced, what is the MOST appropriate evolution of the team's role?

- A. Shift staff from data entry to reviewing exceptions, resolving supplier issues, and overseeing AI extraction quality
- B. Reassign the team to other departments immediately and let the vendor monitor the AI tool's extraction accuracy
- C. Keep staff keying every invoice manually in parallel with the AI indefinitely to cross-check each extracted field
- D. Remove all human involvement and let the AI post every extracted invoice directly to the ledger without review

**Answer:** A

**Explanation:** AI handles routine extraction, while people move to exception handling, judgment-based supplier work, and oversight of AI output quality, which combines human strengths with automation. Removing humans entirely eliminates oversight of errors. Permanent parallel manual entry wastes the investment. Reassigning the team and handing quality monitoring to the vendor loses internal process knowledge and accountability and ignores reskilling.

### Question 10 of 10

A logistics company at an early maturity level wants to become an AI-driven enterprise within three years. Its budget is limited. How should it sequence its development investments?

- A. Buy enterprise licences for a broad set of AI tools in year one so every team can experiment from the start
- B. Launch an autonomous multi-agent supply chain system as the flagship first project to demonstrate ambition to the board
- C. Put most of the budget into an in-house research lab that develops proprietary models before any business use case
- D. Fund data foundations, governance, and skills around a few high-value use cases first, then scale as maturity grows

**Answer:** D

**Explanation:** At early maturity, foundational capabilities and a few value-proving use cases create the base for later expansion and build credibility for further investment. A highly autonomous multi-agent system is too ambitious without basic data and governance. A research lab delays business value on a limited budget, and broad tool licensing without foundations spreads money thinly with little measurable return.

Ready for a timed run? Take the free 15-question AIB-C01 mini exam: https://devopscodelab.com/app/exams/aws-aib-c01/mini/

DevOpsCode Lab is not affiliated with AWS. AIB-C01 and related names are trademarks of their owners. We write every question ourselves; these are not real exam questions.
