Your AI Design Tool Doesn't Have Taste. You Do.

By Kristi Short, CEO · CH2 Solutions

Someone on your product team has already run a prompt through Midjourney or Figma AI or Claude Design. Something came back that looked good. The team shipped it. A few months later, they're wondering why users aren't converting, why the product feels like everything else, why a competitor with a smaller budget somehow looks sharper.

The answer is usually taste. And AI doesn't have any.

What AI design tools actually do

AI generates design by finding patterns in what already exists. Feed it enough Dribbble shots, enough Behance portfolios, enough UI libraries, and it learns to produce work that resembles the average of all of it. For a first draft, a mood board or a quick mockup, that's useful.

However, resembling the average of existing work is the opposite of what design is supposed to do.

Good design makes choices. It decides: this user, this moment, this feeling. That decision requires knowing who the user actually is, what they believe, what will feel right, and what will feel slightly off. None of that lives in training data. It lives in the designer's head after they've done the research, sat with the problem, and made a judgment call.

The gap in UX

UX makes this most obvious, because the whole job is built around a gap: the distance between what users ask for and what they actually need.

Users describe their problems in terms of the solutions they already know. The designer's job is to hear the description and identify the need underneath it. That requires interpretation, understanding context that isn't stated, knowing when the data is misleading and why.

AI works on what's in front of it. A user flow that tests well in a prototype session can fail when it meets a non-predictive real-world user.  A designer who understands the psychology of the human user and has seen the edge cases designs differently.

That knowledge doesn't come from the average of datasets; it comes from being a person who has studied users, created new projects, and understands where user patterns need to be predictive.

What professional oversight actually looks like

AI has a place in the design process. What it doesn't have is responsibility for the decisions it can't make. Someone needs to hold that responsibility. That someone needs to understand design well enough to know when Agentic Design gets it wrong.

In practice, that means three things.

1. Recognizing when work is technically correct but emotionally flat. A color palette can pass every accessibility check and still feel clinical for a product that needs to feel safe. AI won't notice, but a designer will.

2. Knowing when "original" is actually borrowed. AI doesn't cite its influences. A trained eye spots them.

3. Understanding that the brief isn't the whole story. The client says "clean and modern." What they might mean is "trustworthy and not intimidating to a 62-year-old first-time user." The designer has to anticipate what the end user actually needs.

The real risk of removing the professional

Fully AI-generated design won't always be ugly. It will often look fine, but it will look generic in ways that compound over time.

Generic design doesn't fail loudly. It fails quietly, through a product that never builds loyalty, a brand that's hard to remember, a checkout flow that converts at 2.1% instead of 3.4%, and nobody can explain exactly why. That's the designer's domain: the gap between fine and right.

AI gets you to fine design, fast. Getting to RIGHT still takes a human who has taste, knows what they're looking at, and can make the right decisions.

This is a skills argument, not a nostalgia argument

Design is changing. The tools are impressive, and the pace is accelerating. That's not the argument.

The argument is that the people buying AI design tools often don't understand that they're buying speed and volume. However, they're not buying user experience expertise.

The companies that build the best products over the next decade will use AI to move faster, and they will keep designers in the room to make sure faster doesn't mean losing their human audience.

When both pieces work together, everyone wins.

Kristi Short is CEO and Co-Founder of CH2 Solutions, a nearshore engineering firm that builds high-performing technical teams for technology companies. 

→ Learn more at ch2solutions.com