Key Takeaways

1. The Role of AI Designers: Moving Beyond the Interface

• Designers working with AI must understand and influence system prompts, not just UI elements.

• AI-generated outputs can’t always be “designed around”—instead, designers should adjust the system prompt itself to improve quality.

• Testing and iterating on prompts is a core skill, just like wireframing or prototyping.

  
  

2. System Prompts Are Not One-Size-Fits-All

• AI models (e.g., OpenAI’s GPT, Anthropic’s Claude, Meta’s Llama) interpret prompts differently based on how they were trained.

• A system prompt that works well in one model might perform poorly in another.

• Understanding these differences helps designers make better decisions when working across AI platforms.

  
  

3. Collaborating with Engineers on AI Output Quality

Rather than just reporting poor AI results, designers can:

• Ask for the current system prompt and review its structure.

• Experiment with adjustments using AI playgrounds (OpenAI, Anthropic, Vellum, etc.).

• Demonstrate improved prompts to engineers to guide product development.

• This approach creates stronger influence over AI products and ensures designers help shape the output itself.

4. Beyond Single-Prompt Testing: Dynamic Prompting & Model Evaluation

Instead of testing one prompt at a time, designers can:

• Create test cases to evaluate AI responses across multiple conditions.

• Compare different model outputs to determine which is most reliable.

• Use ranking and scoring methods to assess AI performance and refine prompts iteratively.

• These techniques help optimize AI behavior for real-world applications.