Chapter 32 Quiz: Building and Managing AI Teams

Multiple Choice

Question 1. Which AI team role is primarily responsible for taking a model from a Jupyter notebook prototype to a reliable, monitored production system?

a) Data scientist b) ML engineer c) Data analyst d) AI product manager

Question 2. Ravi Mehta identifies his biggest hiring mistake as waiting too long to fill which role?

a) Data engineer b) AI ethics specialist c) AI product manager (translator role) d) ML researcher

Question 3. The "full-stack data scientist" myth is problematic primarily because:

a) Full-stack professionals command higher salaries than specialists b) No single person can excel at data engineering, data science, ML engineering, product management, and ethics review simultaneously c) Full-stack professionals are only effective at large companies d) The term is trademarked and cannot be used in job descriptions

Question 4. In the hub-and-spoke team structure model, the "hub" typically provides which of the following? Select the most complete answer.

a) All data science modeling work for the organization b) Shared infrastructure, standards, governance, and specialized expertise c) Business domain knowledge and stakeholder relationships d) Individual project management and sprint planning for each business unit

Question 5. What is the primary risk of the fully embedded (federated) AI team model?

a) AI professionals are too far from business problems b) Infrastructure costs are higher c) Inconsistent practices, duplicated effort, and weak technical culture across isolated team members d) Too much governance and bureaucracy

Question 6. According to the chapter, what is the recommended approach to sprint planning for data science teams?

a) Use fixed deliverables — "Build a model with 90% accuracy by sprint end" b) Phrase sprint goals as learning objectives — "Determine whether customer browsing data improves churn prediction accuracy" c) Avoid sprint planning entirely, as data science is incompatible with agile methodologies d) Use waterfall planning with 6-month milestones instead of sprints

Question 7. Ravi's upskilling program at Athena included three tiers. The Tier 1 "AI for Everyone" program was:

a) Optional, self-paced, targeting power users who would work hands-on with AI tools b) A 6-week certification program for 200 analysts and planners c) Mandatory, 4-hour online course for all 12,000 employees d) A 2-day instructor-led workshop for 800 managers

Question 8. Which funding model for an AI Center of Excellence centrally funds platform and governance services (free to business units) while charging back consulting and project work?

a) Central funding b) Chargeback c) Hybrid d) Subscription

Question 9. Professor Okonkwo identifies the "translation problem" as the most underrated challenge in enterprise AI. The translation problem refers to:

a) Translating AI documentation into multiple languages for global teams b) Converting data from one format to another across systems c) The gap between the language data scientists use (features, accuracy, distributions) and the language business leaders use (revenue, margins, competitive positioning) d) Translating academic research papers into production code

Question 10. When managing a data science team, negative results (experiments that fail to produce useful models) should be:

a) Hidden from stakeholders to maintain confidence in the AI team b) Counted as project failures in performance reviews c) Documented and communicated as valid contributions to organizational knowledge d) Avoided by only pursuing low-risk projects with guaranteed outcomes

Question 11. A company is deciding between building an AI capability in-house and hiring a consultancy. Which factor most strongly favors building in-house?

a) The company needs the capability within 30 days b) The AI capability is a core source of competitive differentiation c) The company has no existing AI team members d) The AI problem requires a standard, commodity solution

Question 12. The T-shaped professional model recommends hiring people with:

a) Broad, shallow knowledge across all AI disciplines b) Deep expertise in one domain and working knowledge of adjacent domains c) Equal depth in at least three technical disciplines d) Deep business knowledge with no technical skills

Question 13. Which of the following is NOT one of Ravi's six lessons learned from building Athena's AI team?

a) Hire the translator before you hire the tenth data scientist b) Don't skip data engineering c) Always hire researchers first to establish technical credibility d) Diversity is not a nice-to-have

Question 14. According to the chapter, the number one reason AI professionals leave organizations (after compensation) is:

a) Lack of work-life balance b) Poor management c) Boredom — routine work with no new challenges d) Insufficient conference attendance budget

Question 15. A Tier 2 "AI for Managers" workshop should focus on which learning outcome?

a) Teaching managers to write Python code for data analysis b) Enabling managers to build and deploy ML models independently c) Helping managers identify AI opportunities, frame problems, and evaluate AI project proposals d) Training managers to conduct statistical hypothesis tests

Short Answer

Question 16. Ravi's first two data scientists spent approximately 70% of their time on data engineering tasks. Explain why this was problematic from both a cost efficiency and a talent retention perspective. What organizational change resolved the issue?

Question 17. Describe the four functions that distinguish a true AI Center of Excellence from a renamed centralized AI team. Provide one sentence for each function.

Question 18. The chapter identifies five common mistakes in upskilling programs. Choose two of the five mistakes and explain, with a specific example for each, how the mistake would manifest in practice at a company you select (real or hypothetical).

Scenario-Based

Question 19. A financial services company has a centralized AI team of 20 people. Business units report average wait times of 8 weeks from project request to kickoff. Three business units have started hiring their own freelance data scientists without informing the central team. What is happening, and what structural change would you recommend?

Question 20. You are reviewing a consultancy's final deliverable: a customer segmentation model. The model works well in their demo environment. However, when you ask for the training code, they provide an encrypted notebook that runs only on their proprietary platform. What risks does this create, and what contractual provisions should have been in place to prevent this situation?

Question 21. A data science team has been working on a fraud detection model for 6 weeks. The model's precision is high (94%) but recall is low (38%). The data scientists want 4 more weeks to improve recall. The business stakeholder says "94% accuracy is great — deploy it now." Diagnose the communication failure and propose a resolution that both sides would find acceptable.

True or False (with justification)

Question 22. True or false: Most companies need AI/ML researchers on their team to be competitive in AI. Justify your answer in two to three sentences.

Question 23. True or false: Making AI upskilling programs optional is the best approach because it identifies the most motivated employees. Justify your answer in two to three sentences.

Question 24. True or false: The embedded (federated) model, where data scientists report to business unit leaders, is the optimal structure for most organizations. Justify your answer in two to three sentences.

Question 25. True or false: When evaluating AI talent during interviews, traditional whiteboard coding exercises are an effective way to assess data science candidates. Justify your answer in two to three sentences.