Chapter 22 Exercises: No-Code / Low-Code AI

Section A: Recall and Comprehension

Exercise 22.1 Define the following terms in your own words, using no more than two sentences each: (a) no-code AI, (b) low-code AI, (c) AutoML, (d) citizen data scientist, (e) shadow AI.

Exercise 22.2 List the seven steps of a typical AutoML pipeline as described in the chapter (from data ingestion to interpretability analysis). For each step, identify one thing the platform automates and one thing it cannot automate.

Exercise 22.3 The chapter identifies six levels on the AI accessibility spectrum, from "Full Code" to "Prompt-Based AI." Describe each level in one sentence and give one example not mentioned in the chapter for each.

Exercise 22.4 What are the seven dimensions of the vendor evaluation framework? For each dimension, write one evaluation question that the chapter does not explicitly state but that would be valuable for a business leader to ask.

Exercise 22.5 Summarize the three-tier governance model Ravi proposes for Athena's Citizen Data Science Program. For each tier, identify: (a) the type of use case, (b) the governance requirement, and (c) one example from Athena's shadow AI audit that would fall into that tier.

Exercise 22.6 The chapter lists five things AutoML does NOT automate (problem framing, data acquisition, feature understanding, business context for evaluation, deployment). Explain, in your own words, why each of these steps resists automation and requires human judgment.

Exercise 22.7 Distinguish between "build," "buy," and "configure" as AI development strategies. Under what conditions is each approach most appropriate?

Section B: Application

Exercise 22.8: AutoML Platform Comparison Select two AutoML platforms from the following list: DataRobot, H2O Driverless AI, Google Vertex AI AutoML, Azure AutoML, Amazon SageMaker Autopilot. Using vendor documentation, free trials, and published reviews: - (a) Compare the two platforms across the seven evaluation dimensions in the chapter. - (b) Complete the evaluation scorecard with your ratings and justifications. - (c) Recommend one platform for a specific use case (specify the use case). Justify your recommendation in no more than 500 words.

Exercise 22.9: Shadow AI Audit Conduct a shadow AI audit at your current organization (or a previous employer). If you do not have access to an organization, conduct the audit on a fictional company based on published reports of shadow AI incidents. - (a) Identify at least five AI tools or services being used by employees outside formal IT governance. - (b) For each tool, classify the risk level (data leakage, compliance violation, model risk, security vulnerability, or inconsistency) and estimate the severity (low, medium, high). - (c) For each tool, recommend an action: sanction (add to approved list with governance), remediate (address immediate risks while evaluating), or prohibit (too risky without adequate controls). - (d) Draft a one-page executive summary of your audit findings for the CIO.

Exercise 22.10: Citizen Data Science Program Design Design a citizen data science program for one of the following organizations: - (a) A 200-person healthcare technology company - (b) A 5,000-person financial services firm - (c) A 50,000-person global manufacturing company

Your design should include: - An approved tool catalog (at least 5 tools across categories) - A training and certification plan (what topics, how many hours, what assessment) - A tiered governance model (define the tiers, criteria for each, approval processes) - Success metrics (at least 5 quantitative measures) - A risk mitigation plan for the three most likely failure modes

Exercise 22.11: Build vs. Buy vs. Configure Analysis For each of the following Athena Retail Group use cases, recommend whether Athena should build, buy, or configure the solution. Justify each recommendation using the decision matrix from Section 22.12, and identify the key factors that drove your decision. - (a) Product recommendation engine for the e-commerce website - (b) Email send-time optimization for marketing campaigns - (c) Demand forecasting for seasonal inventory planning - (d) Customer sentiment analysis of support tickets - (e) Dynamic pricing for clearance items - (f) Internal document search and Q&A system for employees

Exercise 22.12: No-Code Prototype Using a free trial of an AutoML platform (DataRobot, H2O, or Google Vertex AI AutoML), build a model for one of the following public datasets: - The Kaggle Titanic survival dataset (binary classification) - The UCI Adult Income dataset (binary classification) - The Boston Housing dataset (regression — note: use an ethically appropriate version)

Document your experience: - (a) How long did the process take from data upload to model results? - (b) What was the top model's performance metric? - (c) What features did the platform engineer automatically? Were any surprising? - (d) Could you explain the model's predictions to a non-technical stakeholder? What would you need that the platform did not provide? - (e) What are three things you would want to validate before deploying this model?

Section C: Analysis and Evaluation

Exercise 22.13: The Democratization Debate The chapter presents democratization as both an opportunity and a risk. Write a balanced 800-word essay addressing the following question: Does no-code AI create more value than risk for the average large enterprise?

Your essay should: - Define what you mean by "value" and "risk" in this context - Present at least three arguments in favor of democratization - Present at least three arguments against unrestricted democratization - Propose a framework for determining the appropriate level of democratization for a specific organization - Support your arguments with evidence from the chapter, case studies, or external research

Exercise 22.14: NK vs. Tom — The Deeper Question The opening scene presents NK's two-hour AutoML model (AUC 0.81) against Tom's two-week hand-coded model (AUC 0.83). Analyze this comparison critically: - (a) In what business contexts would NK's approach be clearly superior? Identify at least three scenarios. - (b) In what business contexts would Tom's approach be clearly superior? Identify at least three scenarios. - (c) What information would you need — beyond the AUC score — to determine which model is actually better for Athena's churn prediction use case? - (d) Professor Okonkwo asks whether NK could defend her model in a regulatory proceeding. Under what regulatory frameworks (hint: see the chapter's forward references to Chapters 25-27) might a churn prediction model face regulatory scrutiny? What would a regulator want to see?

Exercise 22.15: The HR Resume Screening Crisis Ravi discovers that Athena's HR team is using an AutoML tool to screen resumes — scoring candidates based on historical hiring data without governance review, bias testing, or legal approval. - (a) Identify at least five specific risks this practice creates for Athena, spanning legal, ethical, reputational, and operational categories. - (b) What types of bias might be present in Athena's historical hiring data? How could these biases manifest in the model's predictions? - (c) Draft a remediation plan for Ravi. What immediate actions should he take? What medium-term actions (3-6 months)? What long-term governance changes should he implement? - (d) If Athena wanted to use AI in hiring responsibly, what safeguards would need to be in place? Reference the tiered governance model from the chapter.

Exercise 22.16: Vendor Lock-In Scenario Analysis Imagine that Athena has been using DataRobot for two years. Fifteen departmental models are running on the platform. DataRobot announces a 40 percent price increase effective in six months. - (a) What is Athena's negotiating position? What factors strengthen or weaken it? - (b) What alternatives does Athena have? For each alternative, identify the switching costs and timeline. - (c) What could Athena have done differently at the outset to reduce its exposure to this scenario? - (d) How does this scenario inform the "Lock-In Risk" dimension of the vendor evaluation framework?

Section D: Integration and Synthesis

Exercise 22.17: Cross-Chapter Integration Draw connections between Chapter 22 and at least three previous chapters: - (a) How does the no-code movement relate to the build-vs-buy framework introduced in Chapter 6? - (b) How does AutoML's automated feature engineering compare to the manual feature engineering you learned in Chapter 7? - (c) How does the citizen data science program connect to the data strategy and governance concepts from Chapter 4? Write a 500-word analysis that synthesizes these connections into a coherent argument about the evolving role of technical skill in AI-driven organizations.

Exercise 22.18: Future Scenario It is 2030. No-code AI platforms can now build models that consistently match or exceed the performance of models built by human data scientists for 90 percent of standard business problems. AutoML handles complex data pipelines, real-time deployment, and automated monitoring. The platforms can even frame business problems based on natural language descriptions. - (a) What role, if any, remains for professional data scientists? What skills become more valuable, and what skills become commoditized? - (b) How does the citizen data science program evolve? What new governance challenges emerge? - (c) What competitive advantage can organizations derive from AI if everyone has access to the same tools? - (d) What does this scenario mean for MBA students studying AI today? What should they prioritize learning?

Selected answers appear in Appendix B: Answers to Selected Exercises.