How does AI change the way we build visualization tools

How does AI change the way we build visualization tools? I believe AI multiplies what’s already there, meaning software design principles matter more than ever. That was the main takeaway from an invited guest lecture I gave last week in Qianwen Wang’s visualization class at the University of Minnesota.

Slides: https://speakerdeck.com/flekschas/data-visualization-tools-in-the-age-of-ai

The goal of a visualization tool is to raise the insight rate, i.e., how quickly a user gets from data to a relational understanding. Many academic visualization tools have practical limitations that hinder adoption (think a steep learning curve or missing practical features), which results in a real-world insight rate that’s not significantly above common plotting tools. So how does AI change this picture?

It depends on what you start from. Applying AI to a closed, monolithic tool only goes so far. You might get a friendlier onboarding step or a chat sidebar but that’s it. Applying AI to a tool that’s already integrated, composable, scalable, and programmable yields compounding benefits: shorter learning curves (AI helps explain and use your tool if it’s programmable), almost no tool switching cost (AI can compose and integrate tools through MCP etc.), and a higher insight ceiling (AI can help surface and explain analysis or data collection artifacts not shown in the vis tool).

Hence, AI multiplies what’s already there! A tool’s engineering design principles are critical as they now sets the ceiling on what AI can do with your tool for your user.

I’m curious how others are thinking about this. Where have you seen AI genuinely compound a vis tool’s usability/usefulness, versus mostly sit on top of one?

And again, many thanks Qianwen Wang for the invite!