GuidesOpinion
The Ring Looks Finished. The Design Isn’t.
AI can generate jewelry images, but will it replace designers? This article argues that the core of design isn't drawing, but judgment and practical skill.
By Timothy Burdairon, Co-founder, Canova
Published

Will AI replace jewelry designers? Not the designers, but a good part of what they have been paid for. The drawing is going to get cheap. The deciding is not, and in a jewelry studio the deciding was always the larger share of the job.
That is a view you can argue with, so here is exactly what I mean. The first image of an idea, the ten variations of it, the render a client needs to see before they believe in a piece: all of that is becoming fast, and the people whose whole value was producing it will feel the squeeze first. The rest of the work stays with someone who knows what happens to metal and stone once the picture is over.
A week at the desk
| Day | Task |
|---|---|
| Monday | Client presents their initial image |
| Tuesday | Clarify aesthetic priorities and practical needs |
| Wednesday | Refine design in CAD for real-world constraints |
| Thursday | Consult with the bench on manufacturing |
| Friday | Deliver the final, wearable piece |
Take an ordinary week. A client arrives with an image they made themselves: an oval stone floating above a fine gold band, an open lattice underneath, every surface polished. They say it is the one. It looks finished, and nothing in it has been checked.

You ask what they love about it. Is it the lattice, or the feeling that the stone hovers? That conversation decides everything after it, and no image answers it for you. Then you learn that they work with their hands and will wear the ring every day. A high setting that catches on gloves and clothing is now a problem, and lowering it too far loses the lightness that sold them.
You are in CAD with the real stone, which has a depth and a girdle that the picture never worried about. The prongs need enough metal to hold it. The join to the band has to survive years of knocks. Take away part of thelattice and the ring is easier to finish but loses its character; add metal underneath and it is sturdier but crowded. You talk to the bench, because the setter has to reach the stone and the polisher has to reach everything else, and a gap that looked elegant on screen turns out to be unreachable.

By the end of the week the ring looks different from the initial image. The client puts it on, and what they see is still what they loved. What they feel is that it fits their hand and their life. The image could not promise either.

Count what in that week an image generator could have done. Mostly the first day's work. Parts of the second, as a way to help a client show what they mean. The rest was judgment about a particular person, a particular stone and a particular bench.
The best argument against
The strongest objection is not that AI is clever. It is that every part of the job above the line is just knowledge, and knowledge can be learned. Proportion for a wearer, setting access, what survives casting: these are patterns, and a system trained on enough finished pieces and enough workshop feedback might absorb them. If the tools also start producing files a workshop can actually use, the gap between picture and object narrows, and clients who arrive with their own image may wonder what they are paying a designer for.
I think that is partly right, and I would not tell a pure renderer or a designer who mostly adapts catalog pieces to feel safe. That work looks like the first to go.
But the objection treats the job as problem solving, and half of it is deciding what the problem is. The client who loves a floating stone does not say they want lightness without height. Someone has to hear it. And when a piece fails on a finger, the person who answers for it is the designer, not the tool. Pattern knowledge may spread widely. Being the one who looked at this client, chose a trade-off and stands behind it does not spread the same way.
So the honest forecast is narrower than either hype or panic. Fewer hours go to drawing. More of the value sits in the questions you ask, the trade-offs you pick, and the conversation with the bench. A designer who only draws has reason to worry. One who decides does not.
Where Canova stands
Canova makes AI tools for the jewelry trade, and we built them for people who do this work, not to stand in for them. Design, currently in private beta by invitation, turns a sketch, a doodle or a CAD view into a realistic render. That is the Monday part of the week. It makes it quicker to show a client a direction. It cannot tell you whether the setting will catch on a glove.
A designer who only draws has reason to worry. One who decides does not.
If you are a designer and someone keeps asking whether you are about to disappear, you can answer plainly. The pictures will come faster. The ring still has to be right.
Sources and method
The information in this article relies on the following sources, which include jewelry industry experts, technical data sheets, and software providers.

