3D configurators are here for good
Last week, “configure your industrial lift and watch it assemble in 3D” was a thing I'd expect from a CPQ vendor's five-year roadmap slide. This week it's a URL on our docs site that works on my phone. We built a proof of concept: a browser-based 3D machine configurator for Autoquip lifts. Pick a product line, set capacity, travel, platform size, and options, and the machine rebuilds itself in front of you — then answer the application questionnaire and your building assembles around it: floors at your measured elevations, the gate types you chose on the sides you chose, enclosure panels, landings. Press and hold the raise button and the carriage climbs the mast while each gate opens as it arrives at that landing — the way real interlocks work — and closes as it leaves. I have not seen anything like that live-from-your-answers experience in industrial automation. Maybe it exists somewhere. If it does, it isn't common, and it certainly isn't a two-day build — except now it is.
What we actually built
Five product lines, three machine archetypes, one config model.
Every model is procedural geometry — parametric three.js, no CAD import. That's the trick that makes it cheap: an approval drawing and the product data sheet give you proportions and ranges; a few hundred lines of geometry code give you a machine that rebuilds at any configuration instantly. The controls behave like the real thing, too: raise and lower are press-and-hold buttons, just like the pendant on the actual lift.
The questionnaire is the configurator. We captured our public VRC application form — about 200 fields across seven steps, with all its conditional branching — into a data-driven question registry. Every product line carries an archetype (scissor / VRC / ram); question sections declare which archetypes they apply to. Common questions (usage, environment, power, finish, engineering) show for every product; the VRC sections (floors and elevation measurements, loading method with full AGV/AMR, conveyor, and fork truck branch questionnaires, containment, mounting, and per-level gates, enclosures, and access sides) appear only for VRC families. Adding a future family means tagging the new line with its archetype: it inherits the whole questionnaire with zero form code.
Answers drive the picture. This is the part that changes the sales conversation. The finish question repaints the machine (pick AQ Safety Yellow, watch it change). The floors answer builds landings at your measured elevations. Gate answers place animated gates of the correct mechanism — swing leaves on hinges, bi-parting pairs, roll-up curtains that roll into their drum, vertical-acting panels riding overhead tracks — on the sides you selected, colored by scope: Autoquip-provided in safety yellow, customer-provided in gray. One button copies a clean application summary — machine configuration plus every answered question — ready to paste into an email or a quote.
The tweaks (a.k.a. the honest part)
Nothing above worked the first time, and the fixes are half the story:
What the customer gets
Walk through it as a buyer:
That last step is the quiet revolution. The gap between “customer describes application” and “everyone sees the same machine” has historically been weeks of back-and-forth, drawings, and imagination. Here it's zero — the specification and the visualization are the same artifact.
Why this is here for good
The economics changed. Procedural 3D means no CAD pipeline, no render farm, no per-model artist time — product data sheets and drawings you already have become geometry parameters. Data-driven questionnaires mean the application engineering knowledge you already encode in web forms becomes conditional logic you write once. A modern browser on a mid-range phone renders all of it at 60fps. And with an AI pair-builder in the loop, the whole proof of concept — five product lines, the questionnaire, the animated site — went from “wouldn't it be cool” to verified-on-production in days, not quarters.
Every industrial manufacturer has the ingredients: product families with parametric variation, application questionnaires, and customers who struggle to picture the thing they're buying. The configurator isn't the future of that conversation. As of this week, it's just how ours works.