Havish M Consulting
Customer Case Study · Manufacturing Robotics
Consumer Durables Manufacturer × Havish M Consulting

Gantry 3D Model

A GenAI-built, browser-based model of the AG-3X gantry turned a spec sheet into a machine customers can spin, explode and jog, on any device.

ClientConsumer Durables Manufacturer FunctionManufacturing Robotics ThemeInteractive 3D Model
Spec sheet to interactive model
0days
▼ vs ~5 weeks for animation
Size of the complete model
0KB
▼ one file, opens on a phone
Components labelled with specs
0
▲ across 5 build stages
Customer demos per month
0
▲ vs 5 with static decks

Figures indicative; validated metrics available on request.

The Challenge

A good machine, hard to show

The robotics team was taking the AG-3X three-axis gantry to customers. The engineering was solid; explaining it was the bottleneck.

CAD stays in CAD

Design files needed licensed software and powerful laptops. Customers and sales could not open them.

Static renders, no motion

Slides showed the gantry, not how X, Y and Z move, or what the gripper does through a pick cycle.

Animation was slow

An outsourced walkthrough took weeks, and every design change meant paying for a re-render.

Training waited for steel

Operators and service teams could only learn the machine once a physical unit existed.

HMC Approach

From spec sheet to spinning model

Engineers supplied the facts. GenAI wrote the 3D code. The result is a single web file anyone can open.

Select a step to see what happened.

Discover · Who needs to see the machine, and what?

We talked to sales, design, service and training about what each wanted to see.

  • Collected the spec sheet, BOM and reference images for the AG-3X.
  • Listed the questions customers ask most: stroke, payload, repeatability, drives and safety.
  • Agreed the must-haves: motion on all three axes, an exploded view and labelled parts.

Deliverable: Model brief and source data pack.

Design · Storyboard the build

The model follows how the machine is built, so it doubles as a teaching tool.

  • Five build stages: frame, motion axes, drive train, end effector, controls.
  • Every label taken from the BOM: part name plus its key spec.
  • Controls chosen to match a real pendant: jog X, Y, Z, gripper and a demo cycle.

Deliverable: Storyboard and component label list.

Build · GenAI writes the 3D code

The model was written as code with Claude, not exported from CAD.

  • Claude generated a single-file WebGL scene with Three.js, iterated from the storyboard.
  • Detail added round by round: T-slot profiles, belts, drag chain, nameplates, PLC and drives.
  • Engineers reviewed proportions and labels against the spec at each round.

Deliverable: AG-3X interactive model, about 80 KB, no install.

Adopt · Put it in front of people

The model went straight into the team's daily work.

  • Sales opens it on a laptop or phone in customer meetings, then shares the link.
  • Design reviews use the exploded view to discuss specific sub-assemblies.
  • Operators and service staff walk through the build stages before the machine arrives.

Deliverable: Demo script and a sharing guide for sales.

Scale · A template for the portfolio

The same pattern now applies to other stations and cells.

  • The AG-3X build becomes the template for future machine models.
  • Labels and specs are kept in step with the BOM as designs change.
  • The robotics team is trained to brief and review new models themselves.

Deliverable: Model template, prompt pack and review checklist.

The Solution

The actual model. Try it.

This is the AG-3X model as delivered. Drag to turn, scroll or pinch to zoom, and use the controls to explode the assembly, jog each axis or run the demo cycle.

Live model · AG-3X Cartesian Gantry · runs in the browser
Loading the 3D model…

What's inside the model

Click a stage to zoom in; click the root to zoom out

AG-3X at a glance

Stroke X × Y × Z
1900 × 1200 × 400 mm
Payload
12 kg
Repeatability
±0.02 mm
X drive
750 W servo, 10:1
Y drive
400 W servo, 5:1
Z drive
400 W, brake
Linear guides
25 mm rails
Frame
80 × 80 T-slot
Drives bus
EtherCAT
End effector
Gripper + 4 vacuum cups

Specifications as labelled in the model.

Impact Dashboard

Faster to make, more people using it

Each kind of machine visual now takes hours, not weeks. And the model reaches far beyond the design team.

Effort per visual asset

Hours to produce, by asset type

BeforeAfter

Who opens the model

Sessions by audience, first 8 weeks

Figures indicative; validated metrics available on request.

Before vs After

How the machine gets explained

DimensionBeforeAfter
Showing the machineStatic renders in a slide deckA live 3D model the customer can turn and zoom
AccessCAD licence and a workstationOne ~80 KB file in any browser, including phones
MotionDescribed in words, or a costly videoJog X, Y, Z and the gripper; watch the demo cycle
Explaining componentsSeparate datasheets per partExploded view with 26 labelled parts and specs
Design changesRe-commission renders or videoUpdate the code with Claude in hours
TrainingOnly once a machine is builtWalk through the build stages before delivery
Tools & Tech Used

No CAD licence required

Claude 3D code generation & iteration Three.js WebGL rendering Spec sheet & BOM source of every label Single HTML file no install, shareable link Any browser desktop, tablet, phone
Key Takeaways

What transfers to any machine builder

01

A link beats a render

Customers engage longer with a machine they can move. A shareable file keeps the conversation going after the meeting.

02

Let the BOM drive the model

Every label came from engineering data. GenAI wrote the code; engineers kept it accurate.

03

Build once, template forever

The first model took the learning. The next machine starts from a working template and a tested prompt pack.