Chapter 35 Exercises: Change Management for AI

Section A: Recall and Comprehension

Exercise 35.1 Define the following terms in your own words, using no more than two sentences each: (a) change management, (b) the ADKAR model, (c) the "last mile" problem, (d) the centaur model, (e) psychological safety.

Exercise 35.2 List the five elements of the ADKAR model in order. For each element, provide one example from the Athena demand forecasting case that illustrates what happens when that element is missing.

Exercise 35.3 Describe the four ways in which AI differs from previous technology waves in terms of change management requirements. Why does each difference matter for organizational adoption?

Exercise 35.4 Explain the McKinsey ratio referenced in the chapter: for every dollar spent on AI technology, three to five dollars should be spent on change management, training, and process redesign. Why is the ratio weighted so heavily toward organizational investment rather than technical investment?

Exercise 35.5 List Kotter's eight steps for leading organizational change. Identify which steps are most commonly skipped in AI transformations and explain why their absence is particularly damaging.

Exercise 35.6 Name the five resistance patterns specific to AI that are identified in the chapter. For each pattern, describe one organizational response that addresses it.

Exercise 35.7 Explain the difference between upskilling and reskilling. Provide one example of each from the Athena case and identify the workforce impact zone (1-4) that each addresses.

Section B: Application

Exercise 35.8: ADKAR Diagnosis Select an AI initiative that has failed or underperformed (either from your own experience, a company you have studied, or a well-known public case such as IBM Watson Health's oncology product).

• (a) For each element of the ADKAR model, assess whether that element was adequately addressed. Use a 1-5 scale (1 = completely absent, 5 = fully achieved).
• (b) Identify the primary bottleneck — the ADKAR element where the breakdown occurred first.
• (c) Propose three specific interventions that would have addressed the bottleneck. For each intervention, specify the timeline, responsible party, and expected outcome.

Exercise 35.9: Resistance Role-Play Analysis You are the VP of Operations at a mid-sized insurance company. Your data science team has built a claims processing model that can approve or flag routine claims in seconds — a task that currently takes adjusters an average of 45 minutes. The model is accurate 91 percent of the time, compared to the adjusters' 88 percent accuracy.

You announce the deployment at a town hall. Three employees stand up:

• Employee A (20-year veteran adjuster): "I've processed over 50,000 claims. No algorithm can understand the nuance I bring to this work."
• Employee B (3-year adjuster): "Are you telling us we're being replaced? Because that's what this sounds like."
• Employee C (team lead): "The model flagged a claim last week that was clearly legitimate. If it can't get that right, why should I trust it?"

For each employee: - (a) Identify which resistance pattern(s) they are exhibiting. - (b) Write a 150-word response that addresses their specific concern without dismissing it. - (c) Describe one structural change (to incentives, processes, or workflows) that would reduce their resistance over time.

Exercise 35.10: Communication Strategy Design An e-commerce company with 5,000 employees is deploying an AI-powered pricing optimization system. The system will recommend price changes for 15,000 SKUs multiple times per day based on demand signals, competitor pricing, and inventory levels. Previously, pricing decisions were made weekly by a team of 12 pricing analysts.

Design a communication strategy by writing the following: - (a) A 200-word email from the CEO to all employees announcing the initiative. - (b) A 300-word briefing document for the 12 pricing analysts explaining how their role will change. - (c) A 100-word FAQ entry for customers who ask "Is an algorithm setting your prices?" - (d) A list of five key messages that should be consistent across all three communications.

Exercise 35.11: Workforce Impact Assessment Consider a hospital system with the following departments and roles. For each role, classify it into one of the four workforce impact zones (Augmented, Restructured, Transitional, Emergent) and justify your classification:

For each role: - (a) Assign a workforce impact zone and explain your reasoning. - (b) Identify the primary resistance pattern you would expect. - (c) Recommend one specific change management intervention.

Exercise 35.12: The Last Mile Audit Select an AI-powered product or feature that you use regularly (a recommendation engine, a smart assistant, a predictive text system, a navigation app, etc.). Evaluate its "last mile" design:

• (a) How seamlessly is the AI integrated into your workflow? Does it require extra steps, or is it embedded in the natural flow of the task?
• (b) Does the system provide feedback loops — i.e., do you see the consequences of following or ignoring its recommendations?
• (c) Can you override the AI's recommendation easily? How does the system respond when you do?
• (d) Based on your analysis, identify one design change that would improve adoption for a less tech-savvy user.

Section C: Analysis and Synthesis

Exercise 35.13: Change Management Plan Design You are the Chief Digital Officer of a national law firm with 800 attorneys and 1,200 support staff. The firm has purchased an AI-powered legal research platform that uses NLP to search case law, generate research memos, and identify relevant precedents. The platform has been tested and performs well — reducing research time by an estimated 40 percent.

However, adoption is your concern. Senior partners (average age 55, average tenure 22 years) are skeptical. Junior associates (average age 29) are cautiously interested but worry that AI will eliminate the research work that traditionally trains new lawyers. Paralegals (who currently do most of the research) are anxious about their roles.

Design a comprehensive change management plan using the frameworks from this chapter. Your plan should include:

• (a) An ADKAR analysis identifying potential gaps for each of the three stakeholder groups.
• (b) A Kotter-aligned organizational strategy with specific actions for each of the eight steps.
• (c) A communication plan with audience-specific messaging for partners, associates, and paralegals.
• (d) A reskilling program design for paralegals, including timeline and format.
• (e) A human-AI collaboration model specifying which decisions the AI makes, which the human makes, and how disagreements are resolved.
• (f) An adoption measurement dashboard with at least five metrics, classified as leading or lagging indicators.
• (g) A plan for sustaining adoption beyond the initial rollout, including governance mechanisms.

This is a comprehensive exercise. Aim for 2,000-3,000 words.

Exercise 35.14: Comparative Framework Analysis Compare and contrast the ADKAR model and Kotter's 8-step model as applied to AI change management:

• (a) What is the unit of analysis for each framework (individual vs. organization)?
• (b) In what situations would you use ADKAR as the primary framework? When would you choose Kotter?
• (c) How can the two frameworks be used together? Provide a specific example showing how ADKAR diagnosis at the individual level informs Kotter actions at the organizational level.
• (d) What are the limitations of each framework when applied specifically to AI — i.e., what aspects of AI change do these frameworks not adequately address?

Exercise 35.15: The Override Dilemma Athena's demand forecasting model achieved a 22 percent override rate after the change management program. Ravi considers this healthy. The COO disagrees and wants to reduce overrides to below 5 percent.

Write a 500-word memo from Ravi to the COO arguing for maintaining the 15-25 percent override rate. Your memo should address:

• (a) Why zero overrides would indicate a different kind of problem (uncritical compliance).
• (b) How the override data improves the model over time.
• (c) How override capability affects adoption and trust.
• (d) What metrics should be used to distinguish "good" overrides (manager has legitimate local knowledge) from "bad" overrides (manager is resisting without justification).

Exercise 35.16: Ethical Dimensions of Change Management The chapter discusses how competitive pressure from NovaMart helped accelerate AI adoption at Athena — employees were more willing to embrace change when the company's survival felt threatened.

• (a) Is it ethical for leaders to use competitive fear as a change management tool? Under what circumstances does urgency-based communication cross the line into manipulation?
• (b) When workforce planning identifies roles in Zone 3 (Transitional), what ethical obligations does the organization have to affected employees? Is providing reskilling and severance sufficient, or does the organization owe something more?
• (c) The chapter describes a case where customer service agents were told the AI was a "co-pilot" when in reality it handled 60 percent of chat interactions and the organization's long-term plan included further automation. At what point does the framing of AI as "augmentation" become dishonest? How should leaders communicate evolving intent?

Section D: Integrative Exercises

Exercise 35.17: Cross-Chapter Integration — From Bias to Trust Trace the connection between the bias crisis in Chapter 25, the governance frameworks of Chapters 27-30, and the trust-building strategies of this chapter. Write a 500-word analysis of how Athena's response to the hiring bias incident either helped or hindered subsequent AI adoption efforts. Reference specific concepts from at least three chapters.

Exercise 35.18: The NovaMart Catalyst The chapter notes that NovaMart's competitive threat (Chapter 37) helped accelerate AI adoption at Athena. Using change management theory:

• (a) Explain which step of Kotter's model this competitive threat primarily supports.
• (b) Identify the risks of relying on external threats as the primary driver of urgency.
• (c) Design an alternative urgency-creation strategy that does not depend on a competitive crisis.

Exercise 35.19: Centaur Design Challenge A commercial bank is deploying an AI model for small business loan decisions. The model analyzes financial statements, cash flow projections, industry risk data, and credit history to produce a loan recommendation (approve, deny, or refer for human review) with a confidence score.

Design the human-AI collaboration workflow:

• (a) Define which decisions the AI should make autonomously, which require human review, and which should always be made by humans. Justify each boundary.
• (b) Describe the information the human loan officer should receive alongside the AI's recommendation (what features were most influential, what comparable loans looked like, etc.).
• (c) Design the override process. What happens when a loan officer disagrees with the AI? Who has final authority? How is the disagreement documented and tracked?
• (d) Identify two scenarios where the AI is likely to be wrong and two scenarios where the human is likely to be wrong. How does your workflow design account for both?

Exercise 35.20: Personal Reflection Reflect on a time when you experienced a technology change in your workplace or academic environment (it does not need to be AI-specific).

• (a) Where did the change management process succeed, and where did it fail?
• (b) Diagnose the failure using the ADKAR model. Which element was the bottleneck?
• (c) What was your personal adoption timeline? Were you an innovator, early adopter, early majority, late majority, or laggard? What influenced your position?
• (d) What would you do differently if you were leading that change initiative, knowing what you now know from this chapter?

Exercises 35.8, 35.13, and 35.19 are recommended for team-based completion. Exercises 35.15 and 35.16 are suggested for in-class discussion. Answers to selected exercises are available in Appendix B.