Why designers call every new tool cheating, from the paintbrush to AI

The Gatekeeper of Your Field

Writer Nicole Alexandra Michaelis published an essay on UX Collective on November 10 describing something I recognize in myself and in much of my field: the sense that every new tool has a "gatekeeper" blocking you at the door.

She cites the case of Ted Chiang, who wrote in The New Yorker that generative AI won't produce art because art is the sum of thousands of small choices, and a machine that makes most of them for you has taken the art out of the process. Michaelis agrees, in part, and then explains how that same distrust kept her away for years from mediums she had always wanted to try.

Her argument isn't that generative AI is good. It's that the discomfort she feels facing the new tool is exactly the same discomfort writers felt facing the typewriter, painters felt facing photography, and monks felt facing the printing press. The question that remains for people who design products, including design tools with built-in AI, is whether we're reacting to what the tool actually does or to the fact that it's new.

Every new medium gets called cheating by the people who are good at the old one.

Nicole Alexandra Michaelis, writer, author of the essay on UX Collective

This Has Happened Before, Always the Same Way

The historical pattern Michaelis gathers is useful because it shows the reaction isn't new, nor exclusive to AI. In 1492, abbot Johannes Trithemius wrote an entire book arguing that monks should keep copying books by hand, because the printing press would make copyists lazy. He had his own book printed.

In 1859, poet Charles Baudelaire declared that photography was where lazy, talentless painters went to hide. Around the same time, painter Paul Delaroche reportedly declared that painting was dead. Photography became a medium of its own, with its own grammar, geniuses, and snobs.

The case that interests me most, as someone who looks at product, is the collapsible paint tube, patented in 1841 by American painter John Goffe Rand. Before it, oil paint was kept in pig bladders tied with string, which practically forced painters to work in a studio. With the tube, paint became portable, painters went outdoors, and Renoir even said that without the tube there would have been no Monet, no Cézanne, and no Impressionism. The tool that lowered the cost of entry didn't destroy painting: it created an entire movement.

What the Research Shows About Who AI Actually Helps

Here's the data point that most deserves the attention of anyone leading a product or design team. Researchers from Stanford and MIT gave more than 5,000 customer support agents access to a generative AI assistant and measured the effect on productivity. The newer, less experienced agents resolved 34% more calls per hour with the tool. Veterans barely changed in performance.

The researchers' explanation is that the assistant transmits the tacit knowledge experienced agents already carried in their heads, and that accelerates the learning curve of people who are just starting out. For anyone who designs products, the direct takeaway is uncomfortable for the expert's ego: generative AI tends to level up those furthest from mastery, not those who already have command of the craft. In design, that means it helps the senior designer less in doing what they already know how to do, and helps the intern more in not getting stuck at the first blank screen.

In short: if the metric that matters on your team is giving fast fluency to people who are starting out, there's real evidence the tool delivers on that. If the metric is replacing the judgment of those who already have a track record, the same research doesn't support that fear.

The Counterargument You Can't Ignore

But there's a point against my own thesis, and it's strong enough to face head-on instead of sweeping under the rug. A 2024 study tested the effect of giving writers AI-generated ideas before they wrote a short story. The result: each individual story was rated as more creative. Except that, looked at together, the stories became more similar to each other.

This is exactly the risk any product designer should take seriously before adopting any layer of generative AI into the workflow. If an entire product category uses the same model, with the same system prompts and the same training examples, the visual and experiential differentiation that today separates one brand from another could shrink. Homogenization isn't an irrational fear from someone fighting the new medium: it's a measured effect, with clear method and limits (a controlled sample of writers, not the entire product design market, but a real signal).

The way out isn't pretending this risk doesn't exist. It's understanding that it lives in how the tool is used, not in the tool itself: whoever uses generative AI just to fill the void of a blank screen risks converging; whoever uses it to explore variation and then applies their own filter of judgment tends to diverge faster than they would alone.

Craft Is the Material's Resistance, Not the Absence of a Tool

Sociologist Richard Sennett describes craft, mastery of a trade, as a conversation with the material: you push, the material pushes back, and it's that resistance that forces revision and, over time, mastery. Wood has grain. Clay collapses if you rush it. Michaelis notes that generative AI also pushes back, just in a more embarrassing way: it does exactly what you asked, not what you meant, and hands it back to you in bullet points.

For people who design interfaces, that friction has practical value: in my reading, a generative AI tool built into a design editor tends to return a generic result on the first try, and the frustration with that result is product data, not a reason to give up. It's the same gap between the idea in your head and what ended up on screen that any designer feels when sketching on paper, before any digital tool exists.

The Question That Remains

Trading your field's gatekeeper for the benefit of the doubt isn't the same as adopting every new tool uncritically. It's recognizing that initial discomfort, by itself, was never proof that something is cheating: historically, it was proof that something was about to change who had access to the medium. The question worth asking before banning or embracing any layer of generative AI in your design workflow isn't "is this dishonest," it's "is this leveling up beginners or replacing the judgment of those who already know what they're doing." Those are different answers, and only one of them justifies fear.