AI design tools sped up designers in one trial, yet some coding trials found slowdowns; what explains the gap?

This explores whether AI design tools such as Figma Make speed people up while AI coding tools slow them down. The corpus suggests the split isn't really design versus coding. A randomized trial of 100 designers and product managers found that Figma Make cut completion time by about 20 percent on structured design tasks Does Figma Make speed up design task completion?. A randomized trial of 96 Google engineers found something similar for code: AI completion, Smart Paste and natural-language-to-code shortened a complex task by about 21 percent, though the confidence interval was wide and statistical significance depended on how the model was specified Do AI coding features actually speed up engineer productivity?. So coding tools can produce speedups too.

The well-known slowdown comes from a narrower setting. Sixteen experienced open-source developers, working on 246 real tasks in mature codebases they already knew deeply, took 19 percent longer with early-2025 AI tools. Before the study, they had predicted a 24 percent speedup Do AI coding tools actually speed up experienced developers?. The reasons the authors give include over-optimism, unreliable AI output, and the developers' own familiarity with their code. When you already know a large codebase well, the AI has less to offer, and checking its suggestions costs you time.

This points to a pattern you might not expect. The design trial found larger gains for product managers than for designers, meaning people working slightly outside their own specialty. The slowdown landed on people working at the center of theirs. Across both domains, the useful variables seem to be how structured the task is and how much the person already knows, not whether the output is a design or a program. Structured tasks done by people with less domain depth give AI the most room to help. Open-ended work on familiar, complex systems gives it the least.

There's also a reason to doubt the stopwatch. Another line of work finds that AI often doesn't reduce total task time so much as move it, away from doing the work and toward writing prompts and checking what comes back Does AI really save time, or just change how we spend it?. A separate randomized trial of developers learning a new library found that AI help weakened conceptual understanding and debugging ability, unless people stayed actively engaged with the material Does AI assistance actually harm the way developers learn?. A 20 percent speedup on a structured design task and a 19 percent slowdown on mature code may both be accurate measurements. Neither tells you what the person learned along the way, or whether the time saved now gets spent later fixing or relearning.

The gaps: the corpus has only one design-tool trial, and it covers structured tasks. Nothing here tests Figma Make on open-ended design work, or with senior designers on design systems they know as well as those developers knew their codebases. That is the comparison that would settle the question.