Starting to build in Revit (and documenting it in public)

I have spent years building inside the Autodesk ecosystem on the manufacturing side: AI agents for Vault, add-ins for Inventor, automation in AutoCAD. That is my home turf. Today I am announcing something different: I am starting to build in Revit, and I will document the process in public, from day one.

No fake expertise

First, the honesty I demand from anyone selling AI applies to me too: I have no previous Revit experience. I will not write expert tutorials or lecture about BIM best practices I have not lived. What I will do is show how someone with years of Inventor and Vault API work learns a new platform, stumbles included.

That format has a name: building in public. And it has an advantage I care about more than the optics: it forces discipline. When you document every week, you cannot sweep the gaps under the rug.

Why Revit, why now

Because the problems I solve do not live in a single product. I already work the AEC side configuring Forma as a common data environment for infrastructure projects. In that world, the Revit model is the central piece: drawings, quantities and coordination all flow from it. Understanding the platform where the model is born makes me better at everything else I already do on that front.

And there is a deeper reason: manufacturing and BIM resemble each other more than their communities admit. A BOM and a quantity schedule tell the same story with different accents. Someone fluent in both sides can move solutions from one world to the other, and that crossing is exactly the space where I like to work.

The same recipe as always

My starting hypothesis is that the recipe that worked for me in Vault and Inventor also applies here. Three steps, in this order.

Learn the API deeply. Before automating anything, understand the data model: what an element is, how families work, what the API exposes and what it does not. In Vault, that groundwork is what later allowed me to build an agent that validates against the real schema instead of hallucinating properties.

Automate the repetitive. Not the glamorous workflows: the clicks that repeat every week. In manufacturing, that ranged from exporting entire libraries to auditing hundreds of components before a release. In Revit, we will see what hurts most: real users will decide that, not my intuition.

Set hard limits. If one day there is an AI agent touching Revit models with my name on it, it will be born with the same defenses as VaultPilot: dry-run, validation, audit trail. That is not negotiable, no matter how exciting the new project feels.

What to expect from this series

I will publish what I find: what feels like Inventor and what feels nothing like it, which parts of the API surprise me, and the first real automation project when it exists. If you work in Revit and have advice for someone arriving from manufacturing, I want to read it. And if there is something you always wanted to automate in Revit, tell me: the best ideas in this series will come from conversations, not from my backlog.