AI vs 3D Rendering Software for Display Concepts

A display manufacturer gets a brief on Monday for a floor unit, three shelves, a header, the client's new range on it. By Wednesday a designer has built it in 3ds Max or SOLIDWORKS, lit it, rendered it, and dropped the render into a store photo. The client looks at it on Thursday and asks for a totally different format. Two and a half days of modeling go in the bin, and nobody bills for them.

I've spent close to four years in point-of-purchase and I build AI POP Displays, so I've watched that week from the manufacturer's side many times. This post is about where AI fits against the 3D rendering stack the industry already runs on. It doesn't replace CAD. It removes the model you were building only to have something to show.

What does the display industry render with today?

The stack is stable and it's good at what it does. As of this month:

• Autodesk 3ds Max 2027, released March 2026, with Arnold as its integrated renderer, at $2,010 a year for one user on an annual plan.
• SOLIDWORKS 2026 with SOLIDWORKS Visualize 2026, where engineered parts get modeled and then rendered. Common wherever metal, wire and timber displays are built.
• KeyShot Studio 2026.2, a dedicated renderer. KeyShot Professional is $1,299 per user per year.
• AutoCAD 2027, also March 2026, for the 2D side, at $2,095 a year for one user.

Board displays add structural packaging CAD on top. Different tools, one shared assumption. Every one of them renders a model somebody built. KeyShot's own import list reads like a directory of CAD formats, from SOLIDWORKS and Rhino to STEP and IGES, because its job begins once the geometry exists.

That assumption is fine after the concept is approved. Before it, it's the most expensive habit in the business.

Why is the concept phase where the hours go?

Because the concept phase is speculative and the modeling isn't. A concept exists to win a yes. Most briefs a manufacturer answers don't turn into orders, and the ones that do rarely survive the first round with their format intact. The client wanted a counter unit, saw it, and now wants a floor stand. Or the brand team changed the hero SKU. Or procurement shopped the brief to three manufacturers and one of them answered first.

Every one of those outcomes throws away a model. The time spent on it was real engineering time, spent on a question that wasn't an engineering question yet. The client isn't asking "how is this built". They're asking "is this the thing we want in the aisle". Building a measured 3D model to answer that is like pouring a foundation to find out whether somebody likes the view.

The render software isn't the problem. It's being used one step too early.

Has the 3D software itself moved toward AI?

Yes, and where the vendors put it is telling.

KeyShot Studio 2026.1 shipped AI Shots. It places your rendered 3D objects into a generated background with correct perspective and scale, running on a model KeyShot says is fine-tuned to work with the CAD data in your scene, and it runs locally. The same release added an edit mode with transform, replace and upscale.

SOLIDWORKS published its AI R2026x FD04 update on 16 September 2026. Among the new features are image generation through natural-language prompts using the Visualize integration, plus AI-generated drawings and design parameterization in natural language.

Autodesk put Autodesk Assistant into both AutoCAD 2027 and 3ds Max 2027 as a tech preview. In AutoCAD it answers questions in context and checks drawings against your own CAD standards. In 3ds Max it finds information about features and workflows inside the product.

Look at the input to each one. AI Shots starts from your 3D objects. The SOLIDWORKS image generation runs through Visualize, which renders a part you already modeled. Autodesk Assistant helps you operate the software. Every vendor pointed its AI at the moment after the model exists, which makes sense for companies whose customers are paying for modeling seats. None of them starts from a client's email.

What does AI change in the concept phase?

It moves the concept image in front of the model instead of behind it.

AI POP Displays starts from the brief. You paste the customer's email and a briefing step reads it and fills the fields a display is decided by: sector, format from a taxonomy where each format carries its proportions, material, style, retail environment. You attach the product shots, the logo, and the customer's sketch or last year's display if there is one. Each reference is labeled by its role and the sketch always goes last, so the model treats a packshot as product to load and a drawing as structure to follow. The concept comes back in under a minute, in a store setting, holding the real packs.

It runs on Google's Nano Banana Pro. The part that decides whether the image is a display rather than a sculpture is the layer on top. A floor unit reads as a floor unit because the format brings its geometry, corrugated reads as corrugated because the material behaves, and the shelves hold what the packs allow rather than what looks full. That's the domain knowledge a designer used to carry into the 3D model by hand, done once in the tool.

The effect on the Monday brief is simple. Three directions go to the client on Monday afternoon instead of one on Thursday, and the client who wanted a floor stand after all says so before anyone has modeled a counter unit. If you have that brief open right now, run it through the free plan and put the result next to the last concept you modeled for it.

Where does the CAD still decide?

Everywhere after the yes. It isn't a small territory.

A concept render has no millimeters in it. It reads as correctly proportioned, and nothing in it should be scaled off. Shelf pitch against the tallest pack, board grade or sheet thickness per part, joints, load on the base, tolerances, dielines, the cut file that goes to the table. All of that is engineering, and it happens in the CAD your team already runs, by people who build displays for a living. The same goes for a final photoreal of an engineered piece for a sell sheet, which is exactly what 3ds Max, Visualize and KeyShot are built to produce.

So the real split is by question, not by tool quality. "Do we want this in the aisle" is a concept question, and an AI concept answers it before anyone opens a modeling seat. "Can we build this, and for how much" is a CAD question, and it starts from the approved concept as the brief. I went through what sits between the two, the six decisions a picture leaves open, in how a brand gets an AI concept built.

How do the two fit into one workflow?

In sequence, with the handoff at approval.

1. Brief in. The customer's email, the packs, any sketch. The sketch-to-render workflow covers how a customer's drawing travels as a reference.
2. Concepts out, same day. Several directions in store context. The client picks, or asks for one change, which costs a credit and about a minute.
3. Approval. The format, the material family and the look are settled.
4. Modeling starts. Now the 3D work is on a piece the client has already said yes to. One model, not three.
5. Engineering, quote, production. Unchanged, in the software you already own.

What drops out is the speculative model in step two, and with it most of the unpaid modeling a manufacturer does to stay in the running. Fifteen free credits cover a real brief end to end, so start with this week's and see how many modeling days it takes out of step two.

Side by side

	3D rendering software (3ds Max, SOLIDWORKS Visualize, KeyShot)	AI POP Displays	Bellto
Starts from	A 3D model someone builds	The client's brief, plus packs, logo and sketch as references	The campaign brief (brand, product, channel, stores, budget, date)
Time to a first concept	Days of modeling, then rendering	Under a minute	Proposes the campaign mix first, then designs each piece with you
Display formats	Whatever you model	Format taxonomy with proportions built in	Floor stand, glorifier, shelf strip, stopper, as a campaign mix
Product load	Modeled by hand	From your real packshots	Proportions derived from the product dimensions and facings
Output	Photoreal render of the modeled piece	In-store concept image a client can approve	3D model, cutlist, dimensioned drawings, STEP, cut DXF per part
AI in 2026	Around an existing model (AI Shots, SOLIDWORKS AI, Autodesk Assistant)	The whole concept step	The whole campaign, brief to shop drawings
Entry	Annual seats, $1,299 to $2,095 a year for the tools above	$0 with 15 one-time credits, Pro $49 founding, list $69	Waitlist, invitation only, no card

What we're building next: Bellto

The table has a third column because the concept isn't where the work ends. After approval, somebody still builds the model by hand, and every late change goes back through the same seat.

Bellto is an AI agent for brands and agencies that takes a POP campaign from brief to shop drawings. You tell it what you're launching, brand, product, channel, stores, budget and date, and it asks for what's missing. It proposes the campaign mix and the materials, floor stand, glorifier, shelf strip, stopper, with the budget split per store. Then it designs every piece with you at real scale in parametric 3D, with proportions derived from the product and the facings, so a dimension change recomputes the whole piece and re-runs the manufacturing checks in seconds. I wrote up what that means on a real piece in parametric POP display design.

Every approved piece comes out as a 3D model, a part-by-part cutlist, dimensioned drawings, a STEP of the assembly and a cut DXF per part, a package a workshop can quote without redrawing. It suggests manufacturers that fit by material, format and volume, and the quote comes from the manufacturer. It's the only route in this comparison that goes from the brief to files a workshop can cut from.

We're building it now. The waitlist is open to any brand or agency, and POP manufacturers can sign up too. We're contacting the first ones soon to run the first real campaigns end to end. It opens by invitation, in small groups, and the campaign you describe when you sign up sets your place. Pricing goes first to the people on the list, and there's no card.

So which one do you use?

• A brief just landed and the client hasn't seen anything yet. AI POP Displays. Get the yes before anybody models.
• The concept is approved and the piece has to be engineered, drawn and quoted. Your CAD and your manufacturer's engineering team, with the approved concept as the brief.
• You need a final photoreal of an engineered piece for a sell sheet. The renderer you already own, fed by the finished model.
• You're on the brand or agency side and the job is a whole campaign, down to files a workshop can cut. Bellto, by invitation from the waitlist.

If the next brief on your desk is a display, sign up free with 15 credits and no card and send the client a concept before you open a modeling seat. Nothing you upload or render trains AI models, on any plan. If the job is the whole campaign, the Bellto waitlist is where it starts.