Chapter 32 Exercises: Building and Managing AI Teams

Section A: AI Roles and Team Composition

Exercise 32.1 — Role Identification

For each of the following tasks, identify the primary role responsible (data engineer, data scientist, ML engineer, AI product manager, data analyst, AI ethics specialist) and explain why that role is best suited for the task. If multiple roles should collaborate, specify the lead and supporting roles.

a) A churn prediction model that performs well in a Jupyter notebook needs to be deployed as a real-time API that integrates with the CRM system. b) Athena's marketing team wants to understand which customer segments responded best to last quarter's promotional campaign. c) The AI team needs to determine whether building a product recommendation system or a demand forecasting model should be prioritized next quarter. d) Customer data from the e-commerce platform, the loyalty program, and the in-store POS system needs to be unified into a single, reliable data pipeline that updates daily. e) A loan approval model needs to be evaluated for disparate impact across racial and gender groups before deployment. f) The VP of Merchandising asks: "Can we predict which new products will succeed?" The request needs to be translated into a precise ML problem statement with defined inputs, outputs, and success metrics.

Exercise 32.2 — Team Composition Design

You are the first AI hire at a mid-size insurance company (5,000 employees, $2 billion revenue). The CEO has given you a budget to hire five AI team members over the next 12 months. The company's priorities are: (1) automate claims processing document review, (2) build a fraud detection model, and (3) improve customer segmentation for marketing.

a) Specify the five roles you would hire, in order, with a brief justification for each prioritization decision. b) For each hire, indicate whether you would recruit externally, upskill internally, or engage a contractor — and why. c) What team structure model (centralized, embedded, hub-and-spoke, CoE) would you recommend at the five-person stage? At what team size would you recommend transitioning to a different model?

Exercise 32.3 — The Full-Stack Trap

You receive the following job posting from a hiring manager:

Data Scientist / ML Engineer / Data Engineer We're looking for a self-starter who can build data pipelines, develop ML models, deploy them to production, manage cloud infrastructure, conduct A/B tests, build dashboards, present to executives, and ensure compliance with data privacy regulations. 5+ years experience required. PhD preferred.

Write a memo to the hiring manager explaining why this job description is problematic. Propose an alternative: either redefine the role with a realistic scope, or propose splitting it into two roles with distinct responsibilities. Be specific about which responsibilities belong to which role.

Section B: Team Structure and Organizational Design

Exercise 32.4 — Org Model Selection

For each of the following organizational profiles, recommend a team structure model (centralized, embedded, hub-and-spoke, or CoE) and justify your recommendation. Consider factors including AI maturity, team size, organizational structure, and business needs.

a) A global pharmaceutical company with AI teams in drug discovery, clinical trials, manufacturing, and commercial. 120 AI professionals across 4 business units. Strong regulatory requirements. b) A 50-person startup with 3 data scientists, no data engineers, and a CEO who wants "AI in everything." c) A regional bank with 8,000 employees, a newly formed data analytics team of 6, and a strategic plan to use AI for credit risk, fraud detection, and customer analytics. d) A consumer electronics company where the engineering division has a mature ML practice (25 people) but the marketing, supply chain, and customer service divisions have no AI capability.

Exercise 32.5 — CoE Charter Development

You have been asked to draft the charter for a new AI Center of Excellence at a retail company with 15,000 employees and an AI team of 30 professionals. The company currently operates a centralized AI team model and is transitioning to a CoE.

Write a one-page CoE charter that includes: a) A mission statement (2-3 sentences) b) A service catalog (at least 5 services with brief descriptions) c) A governance authority statement (what the CoE can approve, require, or block) d) A funding model recommendation with justification e) Three key success metrics with targets for Year 1

Exercise 32.6 — Athena Org Chart Analysis

Review Ravi's team evolution from the chapter opening: - Month 1: 3 people (Ravi + 2 data scientists) - Month 6: 8 people (first ML engineer, first data engineer) - Month 12: 22 people (first AI PM) - Month 18: 35 people (first ethics specialist) - Month 24: 45 people (CoE established)

a) For each stage, describe the most likely organizational pain points (bottlenecks, communication failures, skill gaps). b) At which stage would you have made different hiring decisions than Ravi? Justify your answer. c) Ravi says the AI PM should have been hired at Month 2, not Month 14. If budget constraints limited Month 2 to only 3 people, how would you compose that team differently from Ravi's original trio?

Section C: Talent Strategy

Exercise 32.7 — Hiring Plan

A logistics company is building its first AI team. The VP of Operations has approved a budget for 8 AI hires over 18 months. The company's strategic AI priorities are: (1) route optimization for delivery trucks, (2) demand forecasting for warehouse staffing, and (3) automated package damage detection using computer vision.

Create an 18-month hiring plan that specifies: a) The 8 roles, in hiring order, with start dates b) For each role: job title, key skills, hire externally vs. upskill internally, estimated salary range c) The team structure model at Month 6, Month 12, and Month 18 d) How you would source candidates for the most difficult-to-fill role

Exercise 32.8 — Retention Strategy

You manage an AI team of 12 at a financial services company. Over the past 6 months, you have lost 3 team members to major tech companies (Google, Stripe, and a well-funded AI startup). Exit interviews reveal three consistent themes: (1) compensation gaps — tech companies offered 40-60% more total compensation, (2) lack of career progression — team members felt "stuck" with no clear advancement path, and (3) limited learning opportunities — no conference attendance, no research time, no publication support.

Design a retention strategy that addresses all three themes. Be specific about: a) Compensation adjustments you would propose (considering budget constraints — you cannot match tech company compensation) b) Career path structure (IC and management tracks, with level definitions) c) Learning and development programs (budget, time allocation, specific activities) d) How you would pitch this strategy to the CFO, who views AI team compensation as already "excessive"

Exercise 32.9 — Upskilling Curriculum Design

Design a Tier 2 "AI for Managers" workshop for a healthcare organization with 500 managers across clinical operations, administration, finance, and IT. The workshop should be 2 days (8 hours per day).

Provide: a) A detailed agenda (session titles, durations, and formats — lecture, workshop, group exercise, case discussion) b) Three healthcare-specific case studies you would use (brief descriptions) c) The pre-work you would require participants to complete before the workshop d) How you would measure the workshop's effectiveness at 30 days and 90 days post-completion e) How you would handle resistance from managers who view the workshop as a waste of time

Section D: Cross-Functional Collaboration and Vendor Management

Exercise 32.10 — Translation Exercise

Rewrite each of the following data science statements for a non-technical executive audience. Your rewrite should convey the same information without jargon, include the business implication, and suggest a recommended action.

a) "The model's precision is 92% but its recall is only 64%, which means we're missing a significant number of true positives." b) "We're seeing significant concept drift in the demand forecasting model — the MAE has increased 40% since deployment." c) "The feature importance analysis shows that the top three predictors are recency of last purchase, average order value, and number of support tickets in the last 90 days." d) "We need to retrain the model on the last 6 months of data because the training distribution no longer matches the inference distribution." e) "The model exhibits significant disparate impact on the protected class — the selection rate ratio is 0.62, which falls below the four-fifths rule threshold."

Exercise 32.11 — Vendor Evaluation

A mid-size retailer is considering three options for building a product recommendation engine:

• Option A: In-house build. Hire 2 ML engineers and 1 data scientist. Estimated 6-month build time. Estimated cost: $750,000 first year (salaries + infrastructure), $400,000/year ongoing.
• Option B: SaaS vendor. A recommendation-as-a-service platform. 2-month integration. $200,000/year licensing fee, increasing 10% annually. Limited customization.
• Option C: Consultancy + handoff. An AI consultancy builds the system over 4 months and hands it off to an internal team of 1 ML engineer. Estimated cost: $500,000 consultancy fee + $200,000/year ongoing.

For each option: a) Identify the top 3 risks b) Calculate the 3-year total cost of ownership c) Evaluate strategic fit (assuming the recommendation engine is a potential source of competitive differentiation)

Write a one-page recommendation with your preferred option and justification.

Exercise 32.12 — Knowledge Transfer Plan

Your company has engaged an AI consultancy to build a customer lifetime value (CLV) prediction model. The 4-month engagement is entering Month 3, and you realize there is no formal knowledge transfer plan. The consultancy team of 3 will leave at the end of Month 4, and your internal team (1 data scientist and 1 data engineer) will need to maintain and improve the model.

Design a knowledge transfer plan for the remaining 6 weeks that ensures your internal team can: a) Understand the model architecture and training pipeline b) Retrain the model on new data c) Monitor model performance and respond to drift d) Make incremental improvements to the model

Specify: activities, deliverables, responsible parties, and a timeline.

Section E: Integrated Scenarios

Exercise 32.13 — Startup to Scale-Up

You are the Head of AI at a fast-growing e-commerce company. Over the next 2 years, your AI team needs to grow from 5 people to 40. Revenue is projected to triple. The CEO wants AI embedded in pricing, logistics, marketing personalization, customer service, and fraud detection.

Write a comprehensive 2-year organizational plan that includes: a) Hiring plan (roles, sequence, sources) at 6-month intervals b) Team structure evolution (which model at each stage and when to transition) c) Upskilling plan for the broader organization d) Vendor/partner strategy for capabilities you won't build in-house e) Three key risks to your plan and mitigation strategies for each

Exercise 32.14 — The Broken Team

Read the following scenario and diagnose the problems:

DataPrime Inc. has an AI team of 15 people: 12 data scientists, 2 data engineers, and 1 manager. The team has been in place for 18 months. They have built 23 models — of which only 3 are in production. The business units complain that the AI team "works on whatever they find interesting" rather than business priorities. The data scientists complain that they spend most of their time on data cleaning. The 2 data engineers are overwhelmed. There is no AI product manager. There is no ML engineer. Model deployments are handled by the IT infrastructure team, which has a 6-week deployment queue and limited ML experience. The team sits on a separate floor from the business units. Employee satisfaction surveys show the data scientists rate their "impact on the business" at 2.3 out of 5.

a) Identify at least 5 distinct organizational problems in this scenario. b) For each problem, propose a specific solution. c) Prioritize your solutions: which 2 changes would have the highest impact and should be implemented first? d) Write a 90-day action plan for transforming this team.

Exercise 32.15 — AI Team Budget Justification

The CFO has asked you to justify your proposed AI team budget of $4.2 million for next year (up from $2.8 million this year). The increase covers 4 new hires, a learning budget increase, a new ML platform subscription, and an AI literacy training program for 3,000 employees.

Prepare a one-page budget justification memo that: a) Breaks down the $4.2 million into categories b) Ties each category to a specific business outcome or strategic objective c) Quantifies the expected ROI where possible (refer to the ROI frameworks that will be covered in Chapter 34) d) Addresses the CFO's likely objections (too expensive, unproven value, can't we do more with less?) e) Proposes a "minimum viable budget" alternative if the full request is not approved