As engineering students (or teachers) working on CATIA and the 3DEXPERIENCE platform, you’ve likely wondered: Can AI write my VB.NET macros or automation scripts for me?
The short answer is yes — but with a catch.
If you simply ask a generic Large Language Model (LLM) to "write a CATIA script for ..." you will almost certainly get legacy CATIA V5 code that fails on 3DEXPERIENCE.
Here is a breakdown of why that happens, how to steer AI co-pilots effectively, and what to expect when choosing between different AI models for your CAD automation projects.
1. The V5 vs. 3DEXPERIENCE Trap
Because the web is filled with 20+ years of CATIA V5 documentation and forum posts, AI models default to V5 COM interfaces and syntax. 3DEXPERIENCE (V6) automation patterns are much less common in open training datasets.
When using AI co-pilots for 3DEXPERIENCE:
Set strict guardrails: Explicitly instruct the AI: "Do not use legacy CATIA V5 calls. Focus strictly on 3DEXPERIENCE / V6 automation interfaces."
Provide starting snippets: Feed the model a working 3DEXPERIENCE code template (e.g., how you capture the active window or target a Part in Edit Mode) so it builds on a valid architecture rather than guessing.
2. Treat the AI as a Trainee, Not a Senior Dev
Automation in 3DEXPERIENCE requires precise object navigation and platform-specific logic (like handling PLM session context).
Iterative Guidance: When the model gets stubborn or continuously proposes incorrect method calls, "cut it at the root." Correct its course by pasting specific API error messages or exact parameter structures.
Use Visual Context: If text explanations aren't clicking, upload screenshots of your 3D product structure, parameter tree, or UI state to help the model ground its reasoning.
3. Which AI Model Should You Pick?
Comparing different LLMs (like Google Gemini, Anthropic Claude, and OpenAI GPT) reveals distinct strengths for CAD scripting:
| Model Focus | Strength in CATIA Automation |
|---|---|
| Claude (e.g., Sonnet) | Strong at deep, structural code analysis and step-by-step logic troubleshooting. |
| Gemini | Great for rapid, multi-turn iterative coding and handling multimodal inputs (images/UI screenshots). |
| GPT-4o | Versatile for general VB.NET syntax formatting and refactoring logic. |
Pro Tip: No single model knows 3DEXPERIENCE out of the box without domain context. The best results come from using AI as an interactive partner that you guide with accurate platform rules.
Over to You!
Are you currently using AI tools to write scripts, macros, or apps on the 3DEXPERIENCE platform? What strategies or prompts have worked best for you? Let’s share tips in the comments below!
An article with some screenshots about all these observations can be found herehttps://www.linkedin.com/pulse/ce-model-de-ai-sa-folosesc-pentru-automatizari-catia-fernando-petre-gbubf
