Why this work stays in the shop
Wood model making is a one-off job. Someone hands you a drawing or a sketch, and you turn it into a solid object that fits, measures true and looks right. Software can help plan that object. It cannot choose the board.
Two tasks explain most of the gap. The first is selecting stock: reading grain direction, color and density in a specific plank, then deciding where the part sits in it so the piece does not move, split or blotch under finish. The second is shaping and fitting by hand, where a model maker trims, sands and adjusts a part until it seats against another part the way the drawing intends. Both are judgment calls made with hands and eyes on material that is never twice the same.
The rest of the day keeps the same shape. Setting up a saw or lathe for a short run, cutting reference marks, checking dimensions with calipers and templates, building jigs for an odd shape, repairing a defect that appeared after a cut. That is why the share of task time our method leaves with a person is 94% of the job. The method behind that split is set out in how coverage is measured.
What software does, what it assists, and what it leaves alone
The slice AI handles on its own is small: 0% of task time. It sits in the paperwork around the build. Scaling dimensions from a drawing, converting a sketch into a clean file, and listing the parts and stock a model needs are jobs a computer already does without a person watching.
The assisted slice is about the same size, at 6%. Here a tool speeds up a person who still decides. Nesting software lays out cuts to waste less board. A CNC router can rough a shape from a file. Image tools can show a client three versions of a form before anyone touches a saw. None of that removes the step where the model maker checks the cut part against the template and fixes what is off.
Everything else stays with the worker: choosing and marking the stock, shaping and fitting, finishing and sanding, verifying tolerances, building jigs, and keeping edged tools sharp and machines set. These are the tasks the task list above marks as needing a person, and they are the ones that fill a build day.
What the evidence actually tests
There is no direct head-to-head test of AI against wood model makers yet. Our evidence grade for quality parity is D, which is the grade we use when nothing credible has measured machine output against a qualified professional in this job. Because of that, we publish no parity number here, and you should treat any outside claim that software already matches a trained model maker as untested.
What would settle it is not complicated. Give a shop crew and an automated setup the same drawing, the same stock and the same deadline. Then have experienced shop leads judge the finished models blind on fit, tolerance, surface quality and material use. Until something like that is published, the honest answer is that the physical half of this job has been demonstrated piecemeal, in CNC cutting and sanding of known shapes, and not end to end on fresh work. You can read how we grade evidence on the methodology page.
When this could change
Most likely after 2046 (8 in 10 of our scenarios). For what that window is counting, see how the replacement year is estimated.
Two things could pull it earlier. Cheaper multi-axis routers and sanders would make it worth automating shorter runs than today. And better machine vision for grain and defect detection would chip at the one judgment call that currently has to be made by eye.
Two things hold it back. Our robotics read puts the physical side of the job in fixed automation territory: machines that repeat a programmed path well, not machines that adapt to a knot or a warped board mid-cut. And the money rarely works. Automated setup costs are front-loaded, which suits long production runs; model shops exist to make one of something, then change it.
What to do: if your shop is buying a CNC router, get your name on the programming and setup work rather than only the bench.
How to stay needed
Lean into the parts of the job that stay human. Stock selection is the first: being the person who can look at a pile of boards and say which one will hold a tight tolerance under finish. Hand fitting is the second, especially on assemblies where parts have to mate. Jig and fixture building is the third, because it turns your judgment into something the shop reuses.
Two skills raise the floor. CAD and CAM, so you can take a client’s file straight to a machine and back to the bench, and inspection work, so you can document tolerances rather than just hit them. Both make you the link between the drawing and the finished model.
Shops that cut this work are usually cutting the junior bench seat first, so apprenticeships matter more than they did. Federal data puts this occupation at roughly 280 US jobs with median pay of $56,550, and projects a 3.3% decline over 2025 to 2035 (BLS). It is a small trade, so individual shop closures move the numbers.
If you want to see where the nearby trades land, look at Patternmakers, Wood, Cabinetmakers and Bench Carpenters and Model Makers, Metal and Plastic. You can also browse the wider woodworkers family, the manufacturing sector page, or the list of jobs that mostly need a person. To weigh two trades side by side, use the job comparison tool.