AI Shop Drawings for POP Displays: STEP, Cut DXF and a Cutlist From One Model

A brand marketing manager sends a POP manufacturer a picture of a counter glorifier and asks for a price on 180 stores. Before the shop can reply at all, somebody there has decided the acrylic is 5 mm and not 3, that the base is solvent bonded and not screwed, and that the shelf is 200 mm deep because that is what the render looked like. None of those decisions were in the email. All three of them move the number.

I have spent close to four years in point-of-purchase, and that gap between an approved look and a quotable piece is the one I keep watching campaigns fall into. This post is about what actually closes it, which is a set of files rather than a better picture. What each file is, what the workshop does with it, and what has to be true about the model underneath before any of it is worth sending.

What counts as a shop drawing for a POP display?

In construction, a shop drawing is the detailed version a fabricator draws from the designer's intent, and then someone reviews it. POP works the same way without the vocabulary. The brand approves a look, the shop turns it into something cuttable, and that translation is where the piece is really designed.

For a display, the package that does that job has five parts. A 3D model of the assembly. A part-by-part cutlist. Dimensioned drawings. A STEP of the assembly. One cut DXF per manufactured part. Bellto, an AI agent for POP campaigns that takes a launch from brief to shop drawings, delivers those five formats for every piece you approve, out of the same parametric model.

They are not five views of the same thing. Each one answers a different question at a different desk in the shop, which is why sending four of them is not most of the job.

What does each file actually do at the workshop?

The 3D model. It is the geometry everything else is derived from, and it is also the only file where a human can see the piece as an object. Somebody opens it, rotates it, and confirms that the thing in front of them is the thing the brand approved. If the model is wrong, the four files below are consistently wrong, which is worse than being obviously wrong.

The part-by-part cutlist. One row per part, carrying material, thickness, size and joint. On the four-panel acrylic glorifier, that is four rows for four manufactured parts. That table is what the estimator prices and what purchasing orders sheet against. It is also the shortest answer to "what exactly am I quoting".

The dimensioned drawings. Front, side and top, with measurements on them, and with hidden lines separated onto their own layer. That last detail sounds like housekeeping and is not. Hidden lines are how a bench reads what sits behind what, and if they arrive mixed into the visible geometry the drawing is decorative.

The STEP of the assembly. STEP is the neutral interchange format for solid geometry, so the shop opens the assembly in its own CAD instead of in yours. It can measure anything it wants, check a clearance the drawings did not call out, and drive its own nesting from the same solids.

One cut DXF per manufactured part. Closed contours, one file per part, ready for the cutting table. The glorifier has four manufactured parts and four DXF files. This is the file that turns into motion on a machine, and it is the one with the least room for interpretation.

If a campaign on your desk is currently going out to quote as a picture, the Bellto waitlist is where I would put it.

Why can't a drawing just be traced from the render?

Because the render has no scale to trace. A picture of a glorifier carries proportions that look right, and nothing in it says the base is 200 mm wide or the panel 3 mm thick. A drawing drafted from an image takes its scale from whoever supplies a reference measurement and its geometry from whatever the picture happens to show, so the drawing inherits every guess the render made. The shop then checks those guesses against the real product, finds that a shelf is too shallow for the pack or a joint cannot work at that thickness, and starts again.

A drawing that comes out of a solid model works the other way round. The model holds the dimensions, the parts and the joints, and the drawings, the cutlist, the STEP and the cut files are all views of it. If the pack changes, the model changes, and the views follow without anybody redrawing them.

That is why a set of drawings is not a better picture of the same display. It is a different kind of object, and it is the one the workshop has always had to build for itself, by hand, one quote at a time. Closing that gap from the brand side of the desk, before the email goes out, is what Bellto is for.

What has to be true about the model before the files mean anything?

A package is only as good as the checks behind it, and this is the part that separates a set of exports from something a shop can act on. Here are the checks on the acrylic glorifier.

The envelope comes out exact, at 200 by 200 by 250 mm. Overlap between parts is measured by volume and reads 0.0 mm³ across the six pairs of parts, which is the check that says nothing collides on the bench. The STEP is reimported and verified with no solids lost, so what leaves the model is what arrives in the shop's CAD. The cut DXF files carry closed contours, four of them. And the drawings keep hidden lines on their own layer.

Those five run on the model, not on the picture, and they re-run whenever the model changes. That is the other half of parametric. On the glorifier, a dimension change recomputes the whole piece in under a second. At floor stand scale the same idea holds at a different size. The floor stand example is a four-shelf unit of 1000 by 650 by 1820 mm with a header and casters, 241 parts and 250 cutlist rows, 89 manufactured and 152 bought in. Going from 16 to 20 facings recomputed the entire unit from the parametric geometry alone in 1 min 18 s, with the checks green afterwards.

Why does all of it have to come from one model?

Because five files assembled separately drift, and the drift is invisible until the parts are cut.

If the drawing is drafted by one person and the cut files exported by another, a late change to a panel height reaches one and not the other. The shop cuts to the DXF, the bench assembles to the drawing, and the mismatch shows up as scrap. This is exactly why the cutlist, the drawings, the STEP and the DXF being derived from the same parametric geometry matters more than any single one of them being good. Change the facing count and all five move together, and the checks run again before anything is delivered. Every model goes through engineering review before it goes out.

This is also what makes two quotes comparable for the first time. When every manufacturer receives the same dimensioned piece with the same joints and the same cut files, the difference between two numbers is the difference between two shops rather than between two readings of a photograph. I went through the brand's side of that problem in how to get an AI POP concept built, and the manufacturer's side of the same workflow in where AI fits in a display shop.

It is not a hypothetical shift either. In calls with US POP manufacturers in August 2026, one told me that 4 of roughly 48 proposals in the last six months had started from a client's AI reference, "and more and more". Another said 5 of about 30 orders in the last month did. Brands are already arriving with a picture. The question is what arrives with it.

Where the concept image still fits

The look still has to be approved before anything gets engineered, and a concept render is what does that. AI POP Displays is where the concept image in a POP campaign comes from, and the piece is then designed against it. What changes is the render's job. It stops being the thing you send out to quote and becomes the reference the model is built to.

What we are building: Bellto

Bellto is an AI agent for POP campaigns, and it starts from the campaign rather than from the image. You tell it what you are launching, brand, product, channel, stores, budget and date, and it asks for whatever is missing before proposing anything. Take an illustrative launch, a premium cutlery range going into 180 stores on a 300,000 € budget, which comes to 1,667 € per store. Bellto proposes the campaign mix and the materials, floor stand, glorifier, shelf strip and stopper, with the budget split per store. Then it designs every piece with you at real scale in parametric 3D, with proportions derived from your product dimensions and facings, and a dimension change recomputes the piece and re-runs the manufacturing checks in seconds. Every approved piece comes out in the five formats above, and it suggests manufacturers that fit by material, format and volume. The quote comes from the manufacturer.

We are building it now. The waitlist is open to any brand or agency, and POP manufacturers can sign up too, because what reaches them already carries dimensions, a cutlist and cut files. We are contacting the first ones soon to run the first real campaigns end to end. It opens by invitation, in small groups. Pricing goes first to the people on the list, and there is no card. The full reference page is what Bellto is.

If the next launch on your calendar is one you would rather send out as five files than as one picture, tell us about that campaign on the Bellto waitlist.