Chapter 37 Exercises: Emerging AI Technologies

Chapter 37 Exercises: Emerging AI Technologies

Section A: Recall and Comprehension

Exercise 37.1 Define the following terms in your own words, using no more than two sentences each: (a) agentic AI, (b) multi-agent system, (c) edge AI, (d) small language model, (e) neuromorphic computing, (f) synthetic data, (g) knowledge distillation.

Exercise 37.2 Explain the four rings of the AI Technology Radar (Hold, Assess, Trial, Adopt). For each ring, describe the appropriate level of organizational investment and provide one example technology from the chapter that currently sits in that ring.

Exercise 37.3 List five governance requirements for deploying agentic AI in an enterprise setting. For each requirement, explain in one sentence why it is necessary.

Exercise 37.4 Describe four advantages of edge AI over cloud-based inference. For each advantage, identify one industry or use case where that advantage is most critical.

Exercise 37.5 What is the current state of quantum machine learning? Distinguish between what has been demonstrated in research and what remains theoretical. Explain why the gap between quantum computing theory and hardware matters for business investment decisions.

Exercise 37.6 Compare and contrast the open-weight model approach (e.g., Meta Llama, Mistral) with the closed model approach (e.g., GPT-4, Claude) across at least five dimensions. Under what circumstances would each approach be preferred?

Exercise 37.7 Explain the concept of "model collapse" as described in the Shumailov et al. research. Why does this finding have implications for the future of synthetic data and AI training more broadly?

Section B: Application

Exercise 37.8: AI Technology Radar Select an organization you are familiar with (a current or former employer, a company you have studied, or a well-known public company). Build an AI Technology Radar for that organization by: - (a) Selecting six emerging technologies from this chapter. - (b) For each technology, writing a one-page evaluation that answers the five radar questions: What is it? How mature is it? What is the business case for this organization? What are the risks? What is the recommended action (Hold, Assess, Trial, Adopt)? - (c) Placing each technology on the radar and justifying your placement with specific evidence. - (d) Identifying the single technology that should receive the organization's highest priority and explaining why.

Exercise 37.9: Agentic AI Use Case Analysis Identify three processes in a business you know well that could potentially be handled by an AI agent. For each process: - (a) Describe the current process, including the number of steps, the people involved, and the average time to completion. - (b) Assess which portions of the process an AI agent could handle autonomously and which would require human oversight. - (c) Define the authority boundaries: what actions should the agent be permitted to take without human approval? What actions should require escalation? - (d) Estimate the potential time and cost savings, being explicit about your assumptions. - (e) Identify the top three risks and propose mitigation strategies for each.

Exercise 37.10: Edge AI vs. Cloud Decision A mid-sized retail chain with 200 stores is considering deploying AI-powered computer vision for real-time shelf monitoring. The system needs to detect out-of-stock items, misplaced products, and pricing errors using in-store cameras. - (a) Compare the total cost of ownership for two deployment approaches: (i) streaming video to the cloud for processing and (ii) processing video on edge devices in each store. Include hardware, bandwidth, cloud compute, maintenance, and personnel costs in your estimate. - (b) Analyze the privacy implications of each approach. Reference the privacy-by-design principles from Chapter 29. - (c) Evaluate the reliability implications. What happens when the internet connection goes down under each approach? - (d) Make a recommendation and justify it with specific financial and operational arguments.

Exercise 37.11: Small Model vs. Frontier Model Your company's customer service department handles 50,000 inquiries per day. Currently, inquiries are routed to human agents based on keyword matching, which achieves 72 percent accuracy (meaning 28 percent of inquiries are routed to the wrong department, requiring re-routing).

You are evaluating two options: - Option A: Use a frontier model API (e.g., GPT-4 or Claude) at $0.02 per classification to route inquiries, achieving 97 percent accuracy. - Option B: Fine-tune an open 7B-parameter model on your historical routing data and host it on your own infrastructure, achieving 94 percent accuracy at $0.001 per classification.

(a) Calculate the annual cost of each option.
(b) Estimate the value of the accuracy difference. What is each misrouted inquiry costing the business in customer wait time, agent time, and customer satisfaction?
(c) Consider non-financial factors: data privacy, vendor dependency, customization potential, infrastructure requirements.
(d) Make a recommendation. Under what circumstances would you choose the other option?

Exercise 37.12: Build-or-Wait Decision NovaMart has launched an AI shopping agent that is capturing market share from your e-commerce company. Your CTO proposes building a competing agent, estimating $10 million in development costs and 8 months to launch. Your CFO argues that you should wait 12-18 months for the technology to mature and agent-building platforms to commoditize, reducing costs to an estimated $3-4 million. - (a) What market share might you lose in the 12-18 months of waiting? Make and state your assumptions. - (b) What is the risk that the technology matures in ways that disadvantage early movers (e.g., the first-generation agent platform becomes obsolete)? - (c) What is the risk that waiting allows competitors to establish customer habits and switching costs? - (d) Apply the Technology Radar framework: where should "AI shopping agents" sit on your radar, and what does that placement imply about timing? - (e) Make a recommendation. Justify it with both quantitative and qualitative arguments.

Exercise 37.13: Quantum Computing Investment Memo You are a strategy analyst at a large pharmaceutical company. Your CEO has read an article claiming that quantum computing will "revolutionize drug discovery within five years" and has asked you to prepare a one-page investment recommendation. - (a) Summarize the current state of quantum computing as it relates to drug discovery. Be specific about what has been demonstrated and what remains theoretical. - (b) Evaluate the "within five years" timeline claim. What would need to be true for this timeline to hold? - (c) Recommend a specific level of investment: (i) no investment, (ii) monitoring only (assign one person to track developments), (iii) small research partnership ($1-2M/year with a quantum computing company), or (iv) significant internal investment ($10M+/year in quantum computing capability). - (d) Justify your recommendation with reference to the Technology Radar framework and the chapter's discussion of quantum computing's current state.

Section C: Analysis and Evaluation

Exercise 37.14: The Hype Cycle Pattern The chapter describes a recurring pattern in technology forecasting: bold predictions by technologists, followed by slower-than-expected progress, followed by eventual (but different-from-predicted) impact. - (a) Identify three technologies from the past 20 years that followed this pattern. For each, describe the initial hype, the trough of disillusionment, and the eventual (often unexpected) form in which the technology delivered value. - (b) Select one technology from this chapter that you believe is currently at the "Peak of Inflated Expectations." What does the hype cycle pattern suggest about its likely trajectory? - (c) Select one technology that you believe is in the "Trough of Disillusionment" or approaching the "Plateau of Productivity." What evidence supports your assessment? - (d) What are the limitations of the hype cycle as an analytical framework? When does it mislead?

Exercise 37.15: Open Source vs. Closed: Competitive Dynamics Meta's decision to release Llama as open-weight models is one of the most consequential strategic moves in the AI industry. Analyze this decision from multiple perspectives: - (a) Why does Meta benefit from releasing its models as open-weight, even though competitors can use them? (Consider Meta's business model, talent acquisition, ecosystem dynamics, and competitive positioning against OpenAI and Google.) - (b) What are the risks to Meta from this strategy? - (c) How does the availability of strong open-weight models affect the competitive position of companies that sell proprietary model access (OpenAI, Anthropic)? - (d) How does it affect enterprise customers choosing between open and closed models? - (e) Some critics argue that releasing powerful AI models openly is irresponsible because it enables misuse. Evaluate this argument, drawing on the responsible innovation frameworks from Chapter 30.

Exercise 37.16: Athena's Strategic Decision Athena's decision to build an AI shopping assistant with governance guardrails, rather than racing to match NovaMart's speed to market, represents a specific strategic bet. - (a) What assumptions underlie Grace Chen's claim that Athena can "win the race to trust"? Under what market conditions would this strategy succeed? - (b) Under what conditions would this strategy fail? What if customers do not care about governance guardrails and simply want the fastest, most capable agent? - (c) How does the 6-9 month development timeline affect the strategy? What happens if NovaMart improves its agent during that period? - (d) Evaluate Tom's hiring as a consultant. What are the advantages and risks of using an MBA student (even a technically capable one) for a strategic technology initiative? - (e) Design three measurable success criteria that Athena should use to evaluate its AI shopping assistant after launch.

Exercise 37.17: Hardware Economics and Strategy The chapter describes two opposing trends: rising training costs for frontier models and falling inference costs. - (a) Explain why these trends are not contradictory. How can training costs rise while inference costs fall simultaneously? - (b) What are the strategic implications of rising training costs for: (i) startups hoping to build foundation models, (ii) large enterprises choosing between building and buying AI capability, and (iii) the overall competitive structure of the AI industry? - (c) What are the strategic implications of falling inference costs for: (i) companies building AI-powered products, (ii) the viability of the small language model approach, and (iii) the adoption curve for AI in cost-sensitive industries? - (d) The chapter mentions that GPU shortages created a "hardware barrier to entry" for AI startups. Is this barrier likely to persist, diminish, or grow? Support your argument with evidence from the chapter's discussion of custom AI chips and hardware competition.

Exercise 37.18: The Organizational Readiness Assessment Professor Okonkwo argues that "the skill isn't predicting which technologies will win. It's building the organizational capability to evaluate, pilot, and adopt new technologies systematically." - (a) Describe the two failure modes for emerging technology adoption (chasing every trend vs. waiting too long). For each, identify one real-world company that exemplifies the failure mode. - (b) How does the AI Technology Radar address both failure modes? What are the limitations of the radar approach? - (c) Design a "technology readiness assessment" for your organization that evaluates its ability to adopt emerging AI technologies. Include at least eight dimensions (e.g., technical talent, data infrastructure, governance maturity, leadership support, experimentation culture, vendor management capability, change management capacity, financial flexibility). - (d) Rate your organization (or a hypothetical organization) on each dimension. Identify the two dimensions that most urgently need improvement and propose specific actions.

Section D: Integration and Synthesis

Exercise 37.19: Emerging Technology Impact Matrix Create a matrix that maps the twelve emerging technologies discussed in this chapter against five business functions (marketing, operations, finance, HR, customer service). For each cell, indicate whether the technology's likely impact on that function within the next three years is: high, medium, low, or negligible. Write a paragraph justifying your three most surprising (non-obvious) ratings.

Exercise 37.20: Capstone Preview --- Technology Selection This exercise previews the technology selection component of the capstone project in Chapter 39. Select an industry and a specific organization within that industry. Identify the three emerging technologies from this chapter that would create the most strategic value for that organization over the next five years. For each technology: - (a) Describe the specific use case. - (b) Estimate the investment required (order of magnitude). - (c) Estimate the potential value created (revenue increase, cost reduction, risk reduction, or competitive positioning). - (d) Identify the key risks and dependencies. - (e) Propose a phased implementation plan using the Technology Radar framework.

Selected answers to odd-numbered exercises are available in Appendix B. The AI Technology Radar exercise (37.8) will serve as a foundation for the capstone project in Chapter 39.