Designing with AI, Not for AI

Written By Shiqi Liu

Evaluating how AI tools reshape design workflows through a case study of Lovable.dev

Problem

AI tools promise to accelerate design—but often introduce new complexity.

Designers face challenges such as:

  • Lack of control over generated outputs

  • Difficulty understanding AI logic

  • Overwhelming suggestions and decision fatigue

  • Poor alignment with real-world design workflows

Insight

The problem is not that AI lacks capability — but that it lacks clarity.

Users struggle not because AI is weak, but because:

  • it does not communicate intent clearly

  • it breaks users’ mental models

  • it removes meaningful control

Solution

Instead of redesigning a single interface, this project proposes principles for designing better AI-assisted tools.

Key directions include:

  • Make system state and logic visible

  • Reduce cognitive load through structured guidance

  • Provide clear control and reversibility (e.g., version history)

  • Replace technical language with human-centered communication

  • Balance automation with user agency

Key Findings

Key behavioral patterns identified:

  • Non-designers rely on simple, visible interactions

  • Designers demand deeper control and flexibility

  • Users experience decision fatigue from excessive AI suggestions

  • Terminology significantly impacts usability

Process

This project followed a research-driven design approach:

1. Heuristic Evaluation

  • Evaluated against usability principles

  • Identified major usability issues (e.g., lack of undo, unclear feedback)

2. Competitive Analysis

  • Compared with tools like Firebase and CodeFly

  • Identified patterns like version control and dual-mode interfaces

3. Concept Testing

  • Tested multiple design directions

  • Identified strengths in structure, AI suggestions, and version history

4. Usability Testing

  • Conducted user testing across experience levels

  • Observed behavior patterns and cognitive gaps

What I Learned

Designing AI tools is fundamentally about designing relationships, not interfaces.

Key learnings:

  • AI must support—not replace—user thinking

  • Transparency builds trust

  • Control is more important than automation

  • Different users require different levels of complexity

This project shifted my perspective from designing interfaces to designing interactions between humans and intelligent systems.

Shiqi Liu
Previous

Designing Systems for Human Collaboration