Chapter 9 Exercises: Instructional Prompting and Role Assignment

These exercises develop your instructional precision, verb choice, and role assignment capabilities. They range from isolated skill-building to integrated multi-role challenges. Allow 2–3 hours for the full set.

Part A: Verb Choice and Instructional Precision

Exercise 1: Verb Taxonomy Practice

For each of the following tasks, identify the best verb from the taxonomy in Section 9.2 (generative, analytical, transformative, structural, interrogative, or reasoning). Then write a complete instruction using that verb.

Task A: You want the AI to produce a categorized list of risks in a business plan, organized from highest to lowest severity.

Task B: You want the AI to change a formal report into casual language suitable for a team Slack message.

Task C: You want the AI to make a logical case for implementing a four-day work week at your organization.

Task D: You want the AI to find all the places in a document where the author makes a claim without citing evidence.

Task E: You want the AI to walk you through the implications of a decision step by step, so you understand the reasoning rather than just the conclusion.

For each: name the verb category, select your verb, and write the complete instruction (2–3 sentences minimum).

Exercise 2: The Say-Mean Gap Finder

Each of the following instructions contains a say-mean gap — a difference between what the instruction literally says and what the writer likely means. Identify the gap and rewrite each instruction to close it.

"Make this email more professional."
"Write a clearer explanation of this concept."
"Summarize this report briefly."
"Help me improve this presentation."
"Write something persuasive about our product."
"Give me feedback on this proposal."
"Create a more impactful opening paragraph."
"Make the data section easier to understand."

For each, write: (a) what the gap is, and (b) a revised instruction with the gap closed.

Exercise 3: Instructional Modifier Design

For each of the following output problems, write a single instructional modifier (a short phrase or sentence) that would prevent or correct the problem:

The AI provides a balanced "on one hand / on the other hand" assessment when you specifically wanted a strong argument for one side.
The AI adds a disclaimer at the end of every piece of medical information suggesting the reader consult a physician, but your audience is physicians.
The AI uses technical jargon throughout its explanation despite your audience having no technical background.
The AI consistently produces responses that are 600-800 words when you need 150-word responses.
The AI presents uncertain claims with the same confidence as established facts.
The AI's feedback is consistently mild and qualified, even when you need blunt, direct assessment.
The AI writes in third person when you want first-person perspective.
The AI includes extensive background on topics the audience already knows.

Exercise 4: Sequential Instruction Design

The following task is currently stated as a single vague instruction. Break it into a properly sequenced set of numbered instructions with clear dependencies:

Vague instruction: "Help me build a strong business case for our new product."

This task should be broken into at least five sequential steps that (a) each have a clear instruction with a strong verb, (b) build logically on the output of the previous step, and (c) together produce a complete, well-constructed business case.

Write the five (or more) sequential instructions. For each, specify what input it requires from the previous step and what output it should produce.

Part B: Role Assignment

Exercise 5: Role Archetype Selection

For each professional task below, select the role archetype from Section 9.6 that would produce the most useful output. Explain why that archetype is more useful than a generic instruction.

You need to identify the three most likely failure points in a new employee onboarding program.
You have written a complex explanation of GDPR compliance for small business owners and want to know where you have lost them.
You want to find the weakest arguments in your competitive analysis before presenting it to a skeptical VP.
You want to understand what a specific type of enterprise buyer cares about and whether your pitch addresses those concerns.
You are trying to learn about network security and want guidance to work through the problem yourself rather than be given answers.
You need to improve a technical paper for publication in a peer-reviewed journal.
You want to know whether your project plan has unrealistic timeline assumptions.
You need to explain machine learning to a marketing team without any data science background.

For each: name the archetype, explain why it is the right choice, and write the first two sentences of the role assignment prompt.

Exercise 6: The Role Rotation Exercise

This is the core exercise of the chapter. Take a piece of content you have produced (or create a realistic piece for this exercise) and run it through four different role assignments.

Your content: Choose one of the following (or use your own): - A one-page project proposal - A 300-word persuasive argument for a position you hold - A draft email making an important request - A 5-bullet strategy recommendation

The four roles: 1. Expert Reviewer (someone with deep domain expertise in your field) 2. Devil's Advocate (someone who argues strongly against your position) 3. Target Audience Member (a specific, detailed description of one person the content is meant to persuade) 4. Editorial Reviewer (an editor focused on clarity, structure, and argument strength)

For each role: - Write the full role assignment prompt (using the templates from Section 9.19 as a guide) - Run the prompt (if you have access to an AI tool) or predict what feedback each role would produce - Summarize the key feedback from each role

Synthesis question: What do the four perspectives collectively reveal about your content that any single perspective could not? What would you change based on this exercise?

Exercise 7: Building a Role Assignment for a High-Stakes Task

Identify a high-stakes task in your work — something where getting feedback right matters significantly. This might be: - A client deliverable - A proposal or pitch - A significant decision you need to pressure-test - A communication to a senior stakeholder

Build a complete role assignment prompt for a pre-submission review of this task. Your role assignment should: - Define the specific role with background detail (not just a job title) - Include the role's specific concerns, priorities, and skeptical tendencies - Specify the format of feedback you want (what questions the role should answer) - Include a negative instruction that prevents the AI from defaulting to a more positive or balanced stance than the role would naturally take

Write the complete prompt, then write a reflection on what specific feedback you are hoping to surface that you cannot easily surface yourself.

Exercise 8: System-Level vs. Message-Level Role Assignment

You are working on a complex strategic project that requires you to: - Generate three strategic options (needs a neutral, analytical perspective) - Evaluate those options against financial risk (needs a CFO perspective) - Evaluate them against operational feasibility (needs an operations director perspective) - Make a final recommendation (needs a CEO perspective)

Design a session architecture: 1. What system-level role (if any) would you set at the start? 2. What message-level roles would you assign for each of the four phases? 3. Write the system-level role assignment (if applicable) 4. Write the message-level role assignment for at least two of the four phases

Explain your reasoning: why is each role appropriate for its phase, and why did you make the system-level vs. message-level choice you made?

Part C: The Audience Role Technique

Exercise 9: Building a Specific Audience Persona

For a piece of content you create regularly (or a fictional scenario), build a detailed audience persona for the AI to inhabit. Your persona should include:

Demographic specifics (age, role, industry, experience level)
Contextual specifics (how busy they are, where they will read this, how much time they have)
Knowledge specifics (what they know, what they do not know, what they assume)
Emotional/motivational specifics (what they care most about, what they are worried about, what would earn their trust)
Behavioral specifics (their typical engagement pattern — do they skim? read in detail? delegate to staff?)

After building the persona, write the complete audience role prompt, then use it to evaluate a piece of content you have written.

Reflection: What did the persona reveal about your content that you would not have noticed from your own perspective? What specifically would you change?

Exercise 10: Multi-Audience Comparison

Take a single piece of content — a proposal, an explanation, or a recommendation — and run it through two radically different audience personas.

Design personas that represent genuinely different reader types: - Different expertise levels with the subject matter - Different primary concerns or priorities - Different amounts of time and attention available - Different emotional relationships to the topic (skeptical vs. curious, senior vs. junior, etc.)

After running the content through both personas: 1. What would Persona A say are the biggest gaps or problems? 2. What would Persona B say are the biggest gaps or problems? 3. Are there changes that would improve the content for both audiences? What are they? 4. Are there areas where you cannot satisfy both audiences? How would you choose between them?

Part D: Advanced Integration

Exercise 11: The "Perspective Shift" Full Exercise

Using the Perspective Shift technique from Section 9.16, analyze a real or hypothetical decision from three different stakeholder perspectives.

Your decision: Choose a realistic professional decision — implementing a new process, making a strategic change, launching a new product, restructuring a team, or adopting a new tool.

Stakeholders to analyze: Choose three stakeholders with genuinely different interests in this decision.

Your task: 1. Write the full perspective shift prompt for all three stakeholders 2. For each stakeholder, predict (or generate, if you have AI access) their specific concerns, questions, and priorities 3. Write a synthesis: What do all three agree on? Where are the tensions? What does each stakeholder need to hear to support this decision?

Exercise 12: Role Assignment for Ethical Stress-Testing

You are about to launch a new product feature. Design a role assignment prompt that stress-tests the feature from an ethical perspective — asking the AI to identify potential harms, misuses, or unintended consequences that enthusiastic internal teams might overlook.

Your role assignment should include: - A role that is specifically positioned to identify ethical risks (consider: a consumer advocate, a regulator, a journalist, an ethicist, a user from a vulnerable population) - Instructions that activate genuine critical analysis, not reassurance - Specific questions the role should address - An instruction that prevents the AI from adding "but overall this seems beneficial" qualifications at the end

Reflect: What ethical risks did this exercise surface that your team had not discussed?

Exercise 13: The Conditional Instruction Design

Design a conditional instruction set for a document review scenario. The review should take different actions depending on what the reviewer finds.

Your scenario: You are asking the AI to review customer feedback for potential product issues.

Write a conditional instruction prompt that: - Identifies feedback that mentions a specific type of issue (choose: safety, performance, usability) - Takes different actions depending on severity (critical / moderate / minor) - Follows different output formats based on severity level - Has a clear escalation condition (what triggers "flag this immediately")

Your conditional instructions should use clear if/then structure and produce a useful, organized output regardless of which conditions are met.

Exercise 14: The Instruction Chain for Complex Output

Design a five-step instruction chain for producing a complete competitive analysis of two companies in a market of your choice. Each step should: - Have a clear, specific instruction with a strong verb - Define what output it produces - Specify how that output feeds the next step

Write all five steps as numbered instructions, then write a brief paragraph explaining the logic of your sequencing: why does each step need to precede the next?

Exercise 15: Build a Personal Role Assignment Library

Build a personal role assignment library — a document containing ready-to-use role assignment prompts for your five most common professional review or evaluation tasks.

For each of your five tasks: 1. Name the task and describe when you use it 2. Write the complete role assignment template (using Section 9.19 as a model) 3. Add one specific negative instruction that prevents the most common failure mode for that role in your context

This is a real productivity artifact — save it somewhere accessible. Your goal is that the next time you need any of these roles, you can paste the template and only fill in the specific content, rather than constructing the role from scratch.

Reflection Questions

After completing the exercises, consider:

Which role archetype was hardest to write and why? What does the difficulty reveal about your professional context?
When you ran content through multiple role assignments, which role consistently produced the most useful feedback — and why do you think that is?
Is there a task in your work where you have been avoiding AI assistance because the feedback was generic? Would a role assignment fix this? What would the role be?
What is the ethical question you have about role assignment that this chapter did not fully resolve for you?
Thinking about the difference between asking and instructing: what is the default mode you use most often? What would have to change in how you approach AI interactions to shift more consistently toward instructing?