Workshop 4:
Prototyping Your First AI-Enhanced Workflow
Welcome to Week 4 of our AI implementation journey. This week, we'll focus on building your first functional AI workflow prototype.
Let's turn your workflow maps into real, testable processes that deliver immediate value.

by Haojun See

On The Ground - AI Adoption by Businesses and Individuals
Anatomy of an AI-Enhanced Workflow
Trigger
Event that starts the workflow
This could be a form submission, email receipt, scheduled task, or any other action that initiates the process. Identifying consistent triggers helps automate the workflow.
Input
Data to be processed
The raw information that needs transformation. This may include text documents, spreadsheets, images, or structured data. The quality and format of your input significantly impacts AI performance.
AI Task
Processing by AI tool
The specific action performed by the AI system, such as summarization, classification, content generation, or data extraction. This step should leverage AI's strengths while acknowledging its limitations.
Human Review
Verification and editing
Critical evaluation of AI output by a human expert. This ensures accuracy, addresses edge cases, and maintains quality control. The level of review can be adjusted based on risk tolerance and AI reliability.
Final Format
Deliverable end product
The completed output in its intended format, ready for use. This could be a report, email, presentation, or any other finalized document that meets the end user's needs.
Example: Client sends form → Text block → GPT summarizes → Lawyer reviews → Sends email
Detailed Example: A client submits an intake form with their legal issue (Trigger) → The form data is extracted and formatted as a text prompt (Input) → GPT-4 analyzes the situation and generates a preliminary case assessment (AI Task) → A lawyer reviews the assessment, corrects any misinterpretations, and adds professional insights (Human Review) → The finalized assessment is formatted into a professional email with next steps and sent to the client (Final Format).
This Week's Goal
Build Your First Workflow
Create a usable AI workflow based on your Week 2-3 mapping
Use Real Tools
Implement with ChatGPT, Notion, Docs, Email, or WhatsApp
Test With Real Scenarios
Simulate actual work processes with practical examples
Focus on creating something functional that solves a real problem you face regularly.
Ground Rules for Prototyping
Use dummy or anonymized data
Protect sensitive information during testing phases
Include human checkpoints
Don't over-automate; maintain oversight at critical stages
Define clear parameters
Specify exact inputs, expected AI outputs, and human review points
Remember: Start simple. You can add complexity after proving the basic concept works.
Choosing a Prototype Candidate
Weekly Task
Something you already do regularly
Text-Based
Primarily involves written content
Clear I/O
Has defined inputs and expected outputs
Time-Consuming
Currently takes significant effort to complete
Example: Taking Zoom notes → creating summary email → generating shared checklist
Quick Wins to Try
Meeting Notes
Transform raw notes into structured follow-up summaries with action items
Time saved: 15-30 minutes per meeting
Email Management
Convert long emails into concise client-ready updates with key points highlighted
Time saved: 10-20 minutes per communication
Data Summaries
Turn survey results or feedback into concise management briefs with insights
Time saved: 30-60 minutes per report
Choose tasks that deliver immediate value without requiring complex integration.
Prompt Engineering Review
Role Definition
"You are a [role]..." - Defines the AI's perspective and expertise
Objective Setting
"Your job is to [objective]..." - Clarifies the specific goal
Input Format
"Based on [input format]..." - Describes what you're providing
Output Format
"Please return [output format]..." - Specifies desired result structure
Tone Guidance
"Keep it [tone/style]..." - Sets the appropriate communication style
Clearer prompts produce cleaner outputs, reducing human editing time.
Demo – Building One Prototype Live
Start with Raw Input
Begin with unprocessed Zoom meeting notes containing discussion points and decisions
Apply the Prompt
Create a 3-bullet email summary with client-friendly tone from the notes
Compare Versions
Test 2-3 prompt variations to see how outputs differ in quality and focus
Analyze Results
Discuss what changed between versions and identify needed improvements
Your Turn – Draft Your First Prompt
Choose Document
Select a real work document you regularly process
Write Base Prompt
Create your initial instruction set
Add Components
Specify role, task, format, and tone
Test Output
Run your prompt and evaluate results
Improve Twice
Refine your prompt based on initial results
Refining Through Prompt Variations
Use clarity boosters
  • "Only include items mentioned twice or more"
  • "Use simple language, no jargon"
  • "Limit each point to 12 words max"
Specify priority order
  • "Start with most urgent items"
  • "Order by implementation difficulty"
  • "Group by responsible department"
Add formatting guidance
  • "Bold all action items"
  • "Include a 1-sentence summary at top"
  • "Number key findings sequentially"
Ask yourself: How can I reduce ambiguity in the instructions?
Rebuilding the Workflow With GPT in Mind

Receive input
Note, email, PDF or other document
Clean the input
Copy, edit, prepare for processing
Run prompt in ChatGPT
Process with your engineered prompt
Review and edit
Verify accuracy and make adjustments
Send or share output
Deliver the final product
When It Fails – Troubleshooting Your Prototype
Output Too Vague
Add specific constraints like word count, format requirements, or detail level
Missing Information
Clarify which details matter most and explicitly request their inclusion
Too Robotic
Add instructions like "sound conversational and warm" or provide tone examples
Inconsistent Results
Create a template format the AI must follow with numbered sections
Hands-On Session
Focus on creating something you can actually use in your daily work this week.
Prompt Debrief Discussion
Time allocation
Which step took the most time? How could you streamline it?
Clarity improvements
What made your prompt clearer? What specific wording helped?
Output control
Which type of output was hardest to control? Why?
Efficiency gains
Where did you see the biggest time savings potential?
Mini-Sprint Challenge
1
Finalize Workflow
Complete a working version of your AI-enhanced process
2
Document Input
Copy/paste the exact input you'll process
3
Save Prompts
Record your prompt versions (v1 and v2)
4
Compile Outputs
Save examples of AI-generated results
5
Reflect
Note what's immediately usable in your workflow
Complete this sprint before next week's session to maximize feedback value.
Optional Add-On – Visualize Your Process
Draw Your Workflow
Map the complete process from input to output with all steps
Label Weak Points
Identify areas that need improvement or human intervention
Mark Time Savings
Estimate minutes saved at each workflow stage
Prepare this visualization to share next week for targeted feedback.
Key Lessons Recap
1
Risk Reduction
Prototyping reduces uncertainty before full implementation
2
Clarity Wins
Clear prompts beat complex automation attempts
3
Achievable Gains
You're closer to workflow wins than you realize
Start small, test quickly, and build on what works. Real productivity gains come from iterative improvement.
Weekly Prompt Toolkit
Meeting Summaries
"Summarize this in 3 points with next steps"
Quick Replies
"Rewrite this as a WhatsApp reply"
Task Extraction
"Extract all dates and responsible parties"
Tone Adjustment
"Rephrase to sound polite and firm"
Check-In & Reflection Questions
What did you prototype?
Which workflow did you choose to enhance with AI?
What changed?
How did your outputs evolve from first to final version?
What's next?
Which process will you test with real data?
What's the impact?
How much time could this save weekly?
Take time to document these reflections for our discussion next week.
Contact Us
Need Help With Your Prototype?
We hope this week helped you realize how close AI is to fitting into your daily tasks.
Small wins can lead to big changes in your workflow efficiency.
Get in Touch
We're here to support your AI implementation journey!
See you next week as we build on your prototype and scale your AI workflows!