ITIS 5405 AI for Managers
Module 2 - Generative AI for Individual Work & Managerial
Productivity
2.1 GenAI as a Co-Pilot
From Tool to Thinking Partner
Image source: Stock image
Traditional Tools vs. GenAI
Traditional tools GenAI Is Not Just Another Tool
• Execute specific tasks • Generates content, not just processes data
• Follow clear instructions • Responds to open-ended instructions
• Examples: • Can simulate reasoning and structure ideas
• Excel • Produces plausible outputs
• PowerPoint
• CRM systems
GenAI as a Co-Pilot
• Assists, but does not replace
• Supports thinking and execution
• Requires direction and oversight
• Works best with active human involvement
GenAI primarily augments (not replaces) managerial work
• Automation → replaces tasks
• Augmentation → supports human work
Source: How and Why People Use Gen AI (2025).
What GenAI Can Do Well
1. Generate Content
• Write emails, reports, proposals, and marketing copy
• Create presentations and communication materials
• Draft documentation and internal communications
2. Transform and Refine Content
• Rewrite for tone, clarity, and audience
• Summarize long documents into key insights
• Translate and simplify complex information
3. Support Thinking and Ideation
• Generate ideas and alternatives
• Structure problems and frameworks
• Assist with brainstorming and concept development
What GenAI Can Do Well (continued)
4. Synthesize and Organize Information
• Combine information from multiple sources
• Extract key themes and patterns
• Turn unstructured data into structured outputs
5. Assist with Planning and Workflow
• Break down tasks and create action plans
• Generate meeting agendas and summaries
• Support project and time management
6. Enable Scalable Communication and Personalization
• Tailor messages for different audiences
• Support customer engagement at scale
• Adapt content across channels and contexts
What GenAI Cannot Do Well
1. Guarantee Accuracy
• May produce incorrect or fabricated information
• Does not verify facts
• Outputs can be convincing but wrong
2. Understand Context Deeply
• Lacks full awareness of organizational context
• Cannot fully capture nuanced business situations
• May miss critical constraints or priorities
3. Exercise Judgment and Accountability
• Does not make responsible decisions
• Cannot weigh trade-offs like a manager
• No ownership of outcomes
What GenAI Cannot Do Well (continued)
4. Handle High-Stakes Decisions Reliably
• Not suitable for decisions with significant risk
• May oversimplify complex issues
• Lacks real-world consequences
5. Provide Truly Original or Strategic Insight
• Recombines existing patterns rather than creating deep insight
• Often produces generic or conventional ideas
• Limited ability to generate breakthrough strategies
6. Ensure Consistency and Reliability
• Outputs can vary across prompts
• Sensitive to wording and framing
• Not always stable or predictable
How to Approach Modules 2 & 3?
Module 2: Individual Use Module 3: Organizational Use
• Using GenAI in your own work • Applying GenAI across business functions
• Communication, thinking, productivity • Team and workflow integration
• Prompting and evaluation • Lab focus: Beyond chat — expanded GenAI
• Lab focus: GenAI as a chat interface capabilities
• Deep Research
• Canvas
• Nano Banana, Veo
• NotebookLM
• Integration with other Google tools
Image source: GenAI Landscape (2025).
Why Do We Use Google Gemini?
• Integrated ecosystem (Docs, Sheets, Drive, etc.)
• Demonstrates workflow-level value
• Shows how GenAI fits into real work environments
• Not just a chat tool
Can You Use Other GenAI Tools?
• Module 2:
• Yes — you can use any tool
• ChatGPT, Gemini, Claude, etc.
• Module 3:
• Prefer Gemini
• To better understand ecosystem integration
How You Should Learn by Doing
• Engaging with labs
• Experimenting with prompts
• Evaluating outputs critically
DO NOT
• Passively watch videos without practising
• Copy-paste prompts without thinking
• Accept AI outputs without evaluation
How to Get the Most Value
• Customize prompts to your own work
• Test different approaches
• Compare outputs
• Reflect on strengths and limitations
Use GenAI actively, question it critically, and apply it thoughtfully.
Now, let’s get started!
2.2a Prompt Engineering: Part 1
How to Get Better Results from GenAI
Why Prompting Matters
• GenAI does not “understand” intent automatically
• Output depends on how the task is framed
• Small changes in prompts → large changes in results
Prompting Framework: P.C.T.C.
• Persona: Who should the AI act as? (e.g., "Act as a CFO").
• Context: What business situation is it responding to? (e.g., "We are a mid-sized
firm in a saturated market").
• Task: Define the specific deliverable and format (e.g., "Draft a 3-paragraph
memo").
• Constrain: What limits or rules must it follow? (e.g., "We have a $50k budget").
Prompting: Step-by-Step Improvement
Give me ideas to improve customer retention.
Prompting: Add Persona
Act as a customer experience consultant. Give me ideas to improve customer
retention.
Prompting: Add Context
Act as a customer experience consultant. We are a mid-sized online education
company offering professional certificate programs. Over the past 6 months, repeat
enrollment has dropped, and student satisfaction scores have become less
consistent. Give me ideas to improve customer retention.
Prompting: Add Task
Act as a customer experience consultant. We are a mid-sized online education
company offering professional certificate programs. Over the past 6 months, repeat
enrollment has dropped, and student satisfaction scores have become less
consistent. Draft a short management memo that identifies 4 actionable strategies
to improve customer retention and briefly explains why each strategy may help.
Prompting: Add Constraint
Act as a customer experience consultant. We are a mid-sized online education
company offering professional certificate programs. Over the past 6 months, repeat
enrollment has dropped, and student satisfaction scores have become less
consistent. Draft a short management memo that identifies 4 actionable strategies
to improve customer retention and briefly explains why each strategy may help.
Focus only on strategies that can be implemented within 3 months, require no major
technology investment, and can be executed with a limited budget.
Prompting Is Iterative
• First prompt → draft output
• Refine prompt → better output
• Iterate until useful
Limits of Prompting
• Cannot fix incorrect knowledge
• Cannot replace judgment
• May still produce misleading outputs
So, as a manager, you should understand:
• Prompting = managing AI behavior
• Better prompts → better outputs
• Iteration is essential
• Judgment remains critical
Why Is It Called “Prompt Engineering”?
It’s Not Just Asking Questions It Is Iterative and Experimental
• Prompts shape how GenAI behaves • First prompt → draft output
• Small changes → significantly different • Refine → test → improve
outputs • Requires trial and adjustment
• Requires deliberate design, not casual input
It Requires Judgment and Skill
It Involves Structured Problem-Solving
• Evaluate output quality
• Define goals clearly • Identify weaknesses and gaps
• Translate business needs into instructions • Adjust prompts strategically
• Specify format, constraints, and context
Prompting is not “asking questions” — it is
managing how GenAI behaves.
2.2b Prompt Engineering: Part 2
1. The Shift in Mindset
From Search Engine to Reasoning Engine
• The Old Way (Search): You ask for what already exists. The AI acts as a digital library.
• The New Way (Reasoning): You set parameters for what is possible. The AI acts as a
specialized consultant.
• The Key Pivot: Use the P.C.T.C. Framework to move from generic "Wikipedia-style"
answers to executive-grade work products.
• Persona: Who is the AI? (e.g., Senior Strategist)
• Context: What is the business landscape?
• Task: What is the specific deliverable and format?
• Constraints: What are the boundaries (Budget/Time/Policy)?
2. Meta-Prompting for Accuracy
Reducing "Hallucinations" via Collaboration
• The Problem: AI can be "confident but wrong" when it lacks specific data.
• The Solution: Instead of guessing what the AI needs, ask it to "interview" you first.
• The Workflow:
1. State your high-level goal.
2. Ask: "What 5 questions can I answer to make this advice highly accurate?"
3. Provide the data to ground the AI in your small business reality.
• Takeaway: A 60-second "interview" at the start saves 60 minutes of correcting generic
output later.
3. The Power of Prompt Chaining
• Maintaining "High-Resolution" Detail
• The Trap: "Mega-Prompts" (asking for everything at once) lead to shallow, vague
summaries.
• The Chain: Breaking a large project into smaller, connected steps keeps the output precise.
• The Professional Workflow:
• Step 1: Brainstorm services.
• Step 2: Evaluate via SWOT analysis.
• Step 3: Generate a 7-day implementation table.
• Proof of Value: Chaining ensures the final "Action Plan" is detailed enough to use in a real
Monday morning meeting.
4. The Human in the Loop
You are the Pilot; the AI is the Engine
• Strategic Steering: Constraints (like a $25k budget) are not roadblocks—they are tools for
"steering" the AI toward realistic innovation.
• Iteration is the Goal: Success is rarely found in the first prompt. It is found in the second,
third, and fourth refinements.
• The Professional Advantage: The AI provides the speed and the draft, but the MBA
student provides the judgment, verification, and final decision.
Put P.C.T.C. to Work Today
2.3 Responsible Use of GenAI
From Trust to Critical Judgment
The Core Challenge
• Outputs sound fluent and confident
• Errors are often subtle
• Users tend to over-trust
Key Insight
Confidence ≠ Correctness
DEMO — A Convincing but Flawed Answer
Image source: Petro-Canada
Demo Debrief
What Looks Good? What Could Be Wrong?
• Clear structure • No evidence provided
• Logical arguments • May omit key factors
• Professional tone • May include unsupported claims
• Not context-specific
Common Issues in GenAI Outputs
• Hallucinations (fabricated facts)
• Incomplete answers
• Oversimplification
• Bias or misleading framing
Why Do These Errors Occur?
• Predicts likely text, not truth
• Lacks real understanding
• No real-world accountability
Risk of Blind Trust
• Accepting outputs without verification
• Using AI-generated content directly
• Making decisions based on flawed inputs
A Simple Evaluation Framework
Ask:
1. Is it accurate?
2. Is it relevant?
3. Is anything missing?
4. Does it make sense in context?
Managerial Takeaway
• Treat outputs as drafts, not answers
• Always evaluate and verify
• Combine AI output with human judgment
• Maintain accountability
GenAI outputs are not answers — they are drafts
that must be evaluated.