Exercise 5: Team Workflow Build - Part 1

Mr. DIY Thailand - Generative AI Workshop

Objective: Design and begin executing a complete AI-enhanced workflow for a real Mr. DIY scenario

Tools Required:

• ChatGPT (Plus recommended for Deep Research)
• Claude (Free or Pro)
• Team collaboration space (poster/whiteboard)

---

📖 Context & Scenario

About Mr. DIY Thailand

Mr. DIY is Thailand’s leading home improvement and lifestyle retailer with 1000+ stores nationwide. We serve budget-conscious homeowners and DIY enthusiasts with our “Always Low Prices” positioning. Our product range includes 15,000+ SKUs across hardware, household items, electrical supplies, and more.

Key Facts:

• Revenue: 16.2 billion THB annually
• Stores: 1000+ locations across 77 provinces
• Employees: 11,807 team members
• Transactions: 98.5 million customer interactions (2024)
• Brand Promise: “Always Low Prices” - value-focused, accessible (not luxury, not cheap)

Today’s Scenario

Your team will design and execute a complete AI-enhanced workflow for a real Mr. DIY business challenge. This is your opportunity to apply everything you’ve learned about conversation, context, and strategic automation thinking.

Universal Scenario Design: Each team receives a different business scenario, but all scenarios follow the same workflow principles. Your success depends on systematic thinking, quality control, and collaboration - not domain expertise.

---

🎯 Your Team’s Mission

Design and execute the first 2 steps of an end-to-end AI workflow for your assigned Mr. DIY business scenario.

Team Size: 4-5 people Combined Time: Exercise 5 (45 min) + Exercise 6 (45 min) = 90 minutes total Learning Goal: Practice building a complete workflow with professional-quality outputs (no formal submission required)

---

📋 Team Scenarios

Your team will receive ONE scenario card. Click on your assigned scenario below to access the full details.

Note: All scenarios require 4-5 workflow steps and strategic use of AI tools.

---

🔑 Core Principles Applied

This exercise integrates all three principles:

Principle 1: Automation Comes Last

• You’ll design workflows manually first
• Mark automation potential for future (🟢🟡🔴)
• Don’t automate today - identify what COULD be automated

Principle 2: AI is a Conversation

• Each workflow step requires iteration
• Quality improves through back-and-forth dialogue
• Never accept first AI response

Principle 3: Context Engineering

• Each step needs different context
• Build context progressively
• Better context = better outputs

---

📋 Activity Structure

---

🔄 Starting the Workflow

⚠️ IMPORTANT: Start ONE NEW conversation - then continue through all stages

Unlike previous exercises, this is a CONTINUOUS workflow. Start fresh in ChatGPT or Claude, then keep the same conversation open as you progress through Research → Analysis → Insights.

Why one continuous conversation?

• The AI needs to remember context from earlier stages
• Insights depend on the research and analysis you’ve already done
• This mirrors real-world workflow usage

---

Phase 1: Workflow Design & Planning (20 minutes)

Step 1: Read & Understand Scenario (3 min)

• Read your scenario card together as a team
• Clarify the objective: What’s the final deliverable?
• Identify constraints: Budget, timeline, requirements
• Ask TA if anything is unclear

---

Step 2: Map Workflow Steps (12 min)

On your workflow poster, map out 4-5 major steps

For each step, identify:

1. Step name - What are you doing?
2. Context needed - What information is required?
3. AI tool choice - ChatGPT, Claude, or Deep Research?
4. Conversation flow - How will you prompt AI?
5. Quality checkpoint - How will you verify quality?
6. Output - What do you produce in this step?
7. ⭐ Automation potential - Could this step be automated later?

---

Example Workflow Step:

STEP 2: Campaign Theme Development

├── Context needed:
│   Chiang Mai market research, Mr. DIY brand voice, budget
│
├── AI tool:
│   Claude (long context for synthesis)
│
├── Conversation flow:
│   1. "Act as campaign strategist..." (ROLE+INPUT)
│   2. Refine theme based on local culture
│   3. Check alignment with brand
│
├── Quality checkpoint:
│   Brand voice, budget fit, cultural appropriateness
│
├── Output:
│   Campaign theme with 3 supporting concepts
│
└── Automation potential:
    🟡 MEDIUM - Theme generation could be templated

---

Automation Potential Markers:

• 🟢 HIGH - Repetitive, standardized, good automation candidate
• 🟡 MEDIUM - Some variability, partial automation possible
• 🔴 LOW - Requires judgment, human-led (AI assists only)

---

Step 3: Assign Roles (2 min)

• Who will lead which steps?
• Who will capture outputs?
• Who will check quality?
• Everyone participates - no observers!

---

Step 4: Prepare for Execution (3 min)

• Review your first 2 steps in detail
• What context do you need for Step 1?
• What prompts will you use?
• Which team member starts?

---

Team Check-In (5 minutes)

Each team briefly shares:

• What’s your workflow’s end goal?
• How many steps do you have?
• What’s your first step?
• Any questions or blockers?

Instructor and TA provide quick feedback

---

Phase 2: Execute First 2 Workflow Steps (20 minutes)

Now: Put your workflow into action!

Execution Guidelines for Each Step:

1. Gather Context (2-3 min per step)

• Review what context this step needs
• Pull information from previous step (if Step 2)
• Identify any missing information
• Document context sources

---

2. Design Your Prompt (2-3 min per step)

• Use ROLE + INPUT + REQUIREMENTS + OUTPUT structure
• Include all relevant context
• Be specific about what you need
• Plan for conversation, not single-turn

---

3. Execute Conversation (8-10 min per step)

• Send your initial prompt to the AI

• Review AI response

• Ask follow-up questions:

  • Refine tone
  • Add details
  • Request variations
  • Improve quality

• Iterate 3-5 times minimum

• Don’t accept first response!

---

4. Apply Quality Checkpoint (2-3 min per step)

Use QC framework:

• ✅ Brand alignment: Matches Mr. DIY voice?
• ✅ Accuracy: Facts verified, numbers correct?
• ✅ Actionability: Specific and implementable?
• ✅ Thai market fit: Culturally appropriate?

If quality fails, refine through more conversation

---

5. Capture Output (1-2 min per step)

• Save final output (copy/paste or screenshot)
• Document: What worked well
• Note: What was challenging
• Mark automation potential for this step

---

📊 Progress Tracking

After Step 1 (~10 min):

• ✅ Context gathered
• ✅ Prompt designed and executed
• ✅ Conversation completed (3+ turns)
• ✅ Quality checked
• ✅ Output captured

After Step 2 (~20 min):

• ✅ All of above for Step 2
• ✅ Ready for team check-in
• ✅ Documented learnings

---

💭 Reflection Questions

✏️ Personal notes for your own learning - no submission required

1. Workflow Design Process

What was most challenging about mapping your workflow?

Common challenges teams face:

• ☐ Breaking the problem into discrete steps
• ☐ Identifying what context each step needs
• ☐ Deciding which AI tool to use
• ☐ Defining quality checkpoints
• ☐ Other challenges you encountered

Discussion prompt: How did your team’s diverse perspectives help?

---

2. Tool Selection Strategy

How did you decide which AI tool to use for each step?

Consider:

• ChatGPT strengths: Templates, structured content, lists
• Claude strengths: Long context, synthesis, strategic thinking
• Deep Research strengths: Market intel, competitive analysis, trends

Discussion prompt: What tool selection strategy worked best for your team?

---

3. Principle Integration

Which of the three principles did you use most in this exercise?

Reflect on:

• Did you mark automation potential rather than trying to automate? (Principle 1)
• Did you iterate on AI responses multiple times? (Principle 2)
• Did you build context progressively? (Principle 3)

Discussion prompt: Which principle was most valuable in improving your workflow quality?

---

🎓 Key Takeaways

🎯 PRINCIPLES LEARNED:

All three principles work together in real workflows:

• Automation Last: Design manually, automate strategically later
• Conversation: Each step requires iteration for quality
• Context Engineering: Different steps need different context layers

Workflow Design Insights

• Breaking complex problems into 4-5 clear steps makes AI assistance more effective
• Each workflow step needs its own context, tool choice, and quality checkpoint
• Automation potential varies by step - strategic tasks stay manual, repetitive tasks can automate
• Team collaboration improves workflow design and output quality
• Quality control at each step prevents cascading errors

---

🔄 Reusable Prompts

Workflow Step Template:

STEP [N]: [Step Name]

CONTEXT:
- [List required information]

ROLE: Act as [specific role]

INPUT: [Provide relevant information from previous steps]

REQUIREMENTS:
- [Requirement 1]
- [Requirement 2]
- [Requirement 3]

OUTPUT FORMAT: [Describe desired format]

Context Gathering Checklist:

Before starting each workflow step, verify:
- ☐ What's the input from the previous step?
- ☐ What domain knowledge is needed?
- ☐ What constraints apply (budget, time, brand)?
- ☐ What quality standards must be met?
- ☐ What's the specific output format required?

Quality Verification Framework:

For each workflow output, check:
- ACCURACY: Facts verified? Numbers correct?
- BRAND ALIGNMENT: Matches voice and positioning?
- ACTIONABILITY: Specific enough to implement?
- MARKET FIT: Appropriate for Thai/Mr. DIY context?
- COMPLETENESS: All requirements addressed?

---

✅ Quality Checklist

Before finishing this exercise, verify:

• ☐ Workflow has 4-5 clear steps mapped on poster
• ☐ Each step has context, tool choice, and QC defined
• ☐ Automation potential marked for all steps (🟢🟡🔴)
• ☐ First 2 steps executed with iteration (3+ turns)
• ☐ Quality checkpoints applied to both outputs
• ☐ Team collaborated effectively (everyone participated)
• ☐ Outputs documented and ready to present

---

❓ Need Help?

Running out of time?

Try these approaches:

• Focus on completing 2 steps well rather than rushing through all
• Simplify your workflow - can you combine any steps?
• Ask TA to help you prioritize what’s most important
• Remember: Exercise 6 gives you 45 more minutes

Not sure which AI tool to use?

• Start with one and try it
• If it doesn’t work well, switch to another
• Both ChatGPT and Claude can handle most tasks
• The prompting matters more than the tool choice

Workflow feels too complex?

• Can you break it into simpler steps?
• Are some steps actually sub-tasks of a larger step?
• Focus on major steps, not every small action
• Your TA can help you simplify

Team not collaborating well?

• Assign specific roles (Prompter, Reviewer, Documenter, Quality Checker)
• Rotate who leads each step
• Make sure everyone’s voice is heard
• Ask quieter members: “What do you think about this?”

Output quality not meeting standards?

• Review your context - is something missing?
• Iterate more - don’t accept first response
• Check quality framework systematically
• Ask AI: “How can this be improved?”

TA is here to help! Raise your hand if you need guidance.

---

🔜 What’s Next

Exercise 6 (Block 10, 16:00-16:45):

• Complete remaining workflow steps (3-5 total)
• Apply quality control to all outputs
• Create automation roadmap
• Prepare 2-minute team presentation

---

📦 What You’ll Have by End of Exercise 6

No formal hand-in required - these are for your learning and future reference

Your team will create:

• ✅ Workflow poster with all steps mapped (team collaboration artifact)
• ✅ Sample outputs from each step (save for yourself if useful)
• ✅ Automation roadmap (strategic planning practice)
• ✅ 2-minute presentation to share learnings with peers

TA will circulate to support teams

---

🛠️ Tool Reminders

ChatGPT:

• Best for: Templates, structured content, lists
• Tip: Be specific about format and structure

Claude:

• Best for: Long context, synthesis, strategic thinking
• Tip: Build context progressively through conversation

Deep Research:

• Best for: Market research, competitive analysis, trends
• Tip: Clear research question = better results

---

Ready to build something great!