Idea
Allow an AI agent connected through Revit MCP to assist in creating an existing-conditions BIM model from photographs and other reference information.
Architects frequently begin renovation projects with incomplete documentation. We may have photographs, scans of old drawings, site measurements, aerial images, PDFs, Matterport imagery, point clouds or a limited field survey.
An AI agent should be able to analyze this reference material and construct a preliminary Revit existing-conditions model.
For example, the user could provide:
- Exterior and interior photographs
- Existing floor-plan PDFs
- Historical drawings
- Known dimensions
- Ceiling heights
- Door/window measurements
- Site photographs
- Point-cloud or scan information where available
The agent could then create:
- Levels
- Grids
- Exterior and interior walls
- Floors
- Roofs
- Ceilings
- Doors
- Windows
- Stairs
- Major casework or fixtures
- Rooms
- Basic materials
- Existing-condition elevations
Handling uncertainty is critical
The AI should never silently invent dimensions.
Instead, reconstructed geometry should identify its basis and uncertainty. Elements could be categorized as:
- Verified by dimension
- Derived from drawing
- Estimated from imagery
- Assumed / requires field verification
The agent could also generate an Existing Conditions Verification Report listing dimensions or conditions that should be confirmed during the next site visit.
Example prompt
“Using these 42 site photographs, the existing PDF plan and the dimensions I recorded in the field, create a preliminary existing-conditions model. Use verified dimensions wherever available. Where geometry cannot be determined reliably, make a reasonable provisional assumption and flag it for field verification.”
This would be especially valuable for renovation, historic preservation, tenant improvement and due-diligence projects where architects currently spend substantial time manually rebuilding existing conditions before design can begin.
