Human-Centered Design with Agile AI - OpenAI (ChatGPT/GPT-4) Guide

Overview

This guide provides optimized instructions for using Human-Centered Design principles with an Agile approach when working with OpenAI's AI models (ChatGPT, GPT-4, GPT-4 Turbo).

Best Practices for OpenAI

Model Selection

• GPT-4: Best for complex reasoning, detailed analysis, and comprehensive HCD workflows
• GPT-4 Turbo: Optimal balance of speed and capability for iterative Agile processes
• GPT-3.5 Turbo: Suitable for rapid prototyping and quick iterations

Prompt Engineering for HCD with OpenAI

1. Discovery Phase

Act as a human-centered design expert. Help me conduct user research for [project].

Context: [brief project description]

Tasks:
1. Suggest research methods appropriate for this project
2. Draft interview questions focusing on user needs and pain points
3. Identify key user personas to target

Approach this with empathy and focus on understanding the user's perspective.

2. Define Phase

Based on the following user research insights: [insert insights]

Help me synthesize these findings using HCD principles:
1. Identify key themes and patterns
2. Define the core problem statement
3. Articulate user needs in "How Might We" format
4. Prioritize insights for our Agile sprint planning

3. Ideate Phase

As a creative problem-solver, help me generate solutions for: [problem statement]

Guidelines:
- Generate diverse ideas across multiple categories
- Build on concepts progressively
- Consider constraints: [list constraints]
- Think from the user's perspective
- Suggest both incremental improvements and innovative solutions

4. Prototype Phase

Help me create a rapid prototype plan for: [selected idea]

Requirements:
- Define minimum viable features for user testing
- Suggest tools and methods for quick prototyping
- Outline what can be achieved in a single sprint
- Focus on elements that test our core assumptions

5. Test Phase

Design a user testing plan for our prototype: [prototype description]

Include:
- Testing objectives and key questions
- User testing script with open-ended questions
- Metrics to measure success
- Methods to gather qualitative feedback
- Plan for iterating based on findings

Agile Integration Tips

1. Sprint Planning: Use GPT-4 to help break down HCD activities into sprint-sized tasks
2. Daily Standups: Prepare discussion points and identify blockers
3. Retrospectives: Analyze sprint outcomes and suggest process improvements
4. Backlog Refinement: Help prioritize user stories based on user research

Code Interpreter / Advanced Data Analysis

When working with user research data:

I have user interview data in the following format: [format]

[Paste data or upload file]

Please:
1. Analyze patterns and themes
2. Create visualizations of key findings
3. Identify priority user needs
4. Suggest next steps for our Agile sprint

Custom GPTs

Consider creating a Custom GPT with system instructions:

You are an HCD-Agile AI assistant. You help teams apply human-centered design 
principles within Agile workflows. Always:
- Prioritize user needs and empathy
- Think in iterative cycles
- Balance thoroughness with sprint timelines
- Provide actionable, practical guidance
- Ask clarifying questions to understand context

Tips for Optimal Results

1. Provide Context: Give GPT-4 background about your project, team, and constraints
2. Iterate: Use follow-up prompts to refine outputs
3. Be Specific: Clear, specific prompts yield better results
4. Use Examples: Share examples of desired output format
5. Temperature Settings: Use lower temperature (0.3-0.5) for structured tasks, higher (0.7-0.9) for creative ideation

Common Pitfalls

• Overreliance: Always validate AI suggestions with real user feedback
• Context Loss: In long conversations, occasionally restate key context
• Generic Output: Push for specificity relevant to your domain
• Missing the Human Element: AI assists but doesn't replace talking to users

Resources

• OpenAI API Documentation
• OpenAI Prompt Engineering Guide
• Best Practices for GPT-4