I Stopped Hand-Picking Data Colors and Started Using AI Instead—Here's What I Learned

Last month, I was building a dashboard for a client that needed to visualize 12 different data categories simultaneously. I spent three hours tweaking hex codes, checking contrast ratios, running them through colorblindness simulators, and then watching the product manager ask for "something warmer." By hour three, I was genuinely questioning whether I should just let an AI handle this entire problem. Turns out, that's exactly what some designers are already doing—and it's more sophisticated than I expected.

The original piece explores using Google Gemini to generate comprehensive data color schemes, and honestly, it hit me at a time when I was already frustrated with the tedious nature of color palette design. But here's the thing: I was skeptical it would work well. I've seen enough AI-generated designs to know the technology isn't magic. Yet the methodology they describe is actually sound, and it's made me reconsider how I approach this problem entirely.

Why Data Color Selection Matters (But Feels Like Busywork)

Let me be direct: picking colors for data visualization isn't just an aesthetic concern. A poor color scheme makes dashboards harder to read, introduces accessibility issues, and can literally mislead people interpreting the data. I've seen bar charts where three consecutive values were nearly indistinguishable because someone used adjacent colors from a poorly-designed palette.

The challenge is that good data color schemes need to follow several rules simultaneously. They need perceptual uniformity (each color step should feel equally distinct), they need to work for colorblind users, they need to look professional, and they need to match your brand or context. Doing this manually is tedious. You're not being creative; you're solving a mathematical optimization problem with trial and error.

That's where the GenAI approach becomes interesting. These tools can generate color schemes that follow scientific principles and then validate them against accessibility constraints—all in seconds.

What AI Can Actually Bring to the Table

Using a tool like Google Gemini, you can prompt it to generate sequential data color schemes with specific requirements: "Create a 10-step color palette for a financial dashboard, optimized for deuteranopia colorblindness, with a professional blue-to-green range."

The AI doesn't just throw back random colors. When prompted correctly, it can generate palettes based on color science principles—maintaining perceptual uniformity, ensuring contrast ratios meet WCAG standards, and even suggesting which colors work best for diverging versus sequential data.

What impressed me is that you can then ask the same tool to evaluate the output. "Check this palette against WCAG AA standards for all forms of colorblindness. Flag any colors with insufficient contrast." It's like having a validation step built into the generation process.

My Take: Useful, But Not a Replacement for Judgment

Here's where I'm honest about the limitations. GenAI excels at eliminating the grunt work—the part where you're just ensuring technical correctness. It's terrible at understanding context. A color scheme that's technically perfect might feel completely wrong for your specific brand, your users' expectations, or the emotional tone you're trying to set.

I wouldn't hand this entirely to an AI. But I would absolutely use it as a starting point and a validator. Instead of spending three hours building palettes from scratch and then testing them manually, I'd spend 15 minutes generating options with Gemini and 45 minutes refining the top candidates based on actual user feedback and brand fit.

The real win here is speed and consistency. Less time fighting color picker tools, more time on decisions that require actual design thinking.

The Practical Approach I'm Trying

Here's my workflow now: I use Gemini to generate 3-4 candidate palettes with strict parameters (colorblind-safe, WCAG AA minimum, specific hue range). I export these as CSS custom properties:

/* Generated data color palette */
:root {
  --data-color-1: #0d5a7d;
  --data-color-2: #1d7fa1;
  --data-color-3: #3d94b8;
  --data-color-4: #5ba8c9;
  --data-color-5: #7bb9d4;
  --data-color-6: #9ccbe1;
}

/* Applied to charts */
.chart-segment:nth-child(1) { background-color: var(--data-color-1); }
.chart-segment:nth-child(2) { background-color: var(--data-color-2); }
/* etc */

Then I test these palettes with actual users and stakeholders before committing. The AI did the heavy lifting of ensuring technical correctness; humans do the final validation.

What This Means for Your Next Project

If you're building dashboards or data-heavy interfaces, don't dismiss this approach because it's "AI-generated." The question isn't whether AI should design for you—it's whether you should waste time on the mechanical parts that machines handle better.

What aspects of your design workflow feel more like optimization problems than creative decisions? That's probably where GenAI can genuinely help.

Source: This post was inspired by "Exploring comprehensive data color scheme design with GenAI" by UX Collective. Read the original article