What AI Is Good at in UX—and What It Isn’t

A clear-eyed look at where AI actually helps.

In 2026, the question is no longer "Do you use AI?" The question is "Who is driving?"

There is a misconception in the market that AI tools like Midjourney, Lovable, or Galileo are "Designers." They are not. They are high-speed manufacturing plants. If you feed them bad blueprints, they will manufacture garbage at record speeds.

At Proto UX, we embrace an AI-native workflow, but we draw a sharp red line between Execution (The Machine) and Strategy (The Human).

Here is the honest breakdown of where AI creates leverage, and where it hits a hard wall.

What AI is Incredible At (Velocity & Syntax)

We use AI aggressively in these three areas to save our clients time and money:

1. The "Boilerplate" Lift

Building a login screen, a settings panel, or a standard data table is no longer a design challenge. It is a commodity.

• •The AI Role: We use tools to instantly generate the "Standard UI" code and layout.
• •The Benefit: We don't bill you 40 hours for drawing standard inputs. We spend that budget on the 10% of the product that is unique to your business.

2. Pattern Synthesis (Research)

When we conduct 20 hours of stakeholder interviews, the sheer volume of data is overwhelming.

• •The AI Role: AI excels at pattern recognition. We feed it transcripts and ask it to find "The top 3 friction points mentioned by the Sales team vs. the Engineering team."
• •The Benefit: It turns raw noise into structured data instantly, allowing us to move to diagnosis faster.

3. Translation (Design to Code)

This is the superpower. AI bridges the gap between Figma variables and React props.

• •The AI Role: Writing the mundane CSS, ensuring accessibility tags are present, and refactoring components to match the design system.
• •The Benefit: We ship "Dev-Ready" code, not just static pictures.

What AI is Terrible At (Context & Intent)

This is why you pay for a Logic Architect. AI models are Probabilistic Engines—they guess the next likely pixel based on the average of the internet. They do not understand truth.

1. "The Why" (Intent)

AI can design a beautiful dashboard. It cannot tell you if a dashboard is the right solution for the user's problem.

• •The Human Factor: Maybe the user doesn't need a dashboard. Maybe they need a morning email summary. AI will never suggest that because you asked for a dashboard. We define the solution; AI only executes the request.

2. Organizational Physics

AI designs in a vacuum. It assumes infinite budget, perfect data, and modern tech stacks.

• •The Human Factor: Your business has legacy code, specific compliance needs (HIPAA/SOC2), and a unique brand voice. We filter the AI's output through the reality of your constraints.

3. Novel Logic

AI is a remixer. It can only produce variations of things that already exist.

• •The Human Factor: If your product requires a novel interaction model—something that hasn't been done a thousand times on Dribbble—AI will hallucinate broken logic. Complex state management and edge-case error handling require a human architect to map the decision tree.

The Verdict: The Pilot and The Plane

We view AI as the ultimate Force Multiplier.

It allows us to be 10x faster at production. It allows us to be 10x cheaper on the "boring stuff."

But it makes the Human Factors Ph.D. more valuable, not less. Because when you can build anything in seconds, the cost of building the wrong thing goes up exponentially.

We use AI to run the engine. We use Human Intelligence to steer the ship.

Written by Young Ryu, Ph.D.
Principal at Proto UX