Let the AI propose classification codes for elements that arrived without one.
Results are grouped by classification code. When elements come out of the authoring tool without one — a common gap on early models — AI Classification proposes a code for each, and you review every proposal before anything is applied.Open it from Mapping tools on the Materials page.
Pick the building components to classify. The AI suggests a classification code for every model element of a picked component. Only unclassified hides anything that already has a code, which is usually what you want.
One row per element type, showing the category, type and material identifiers, any existing code, and what the AI proposes.
Column
What it tells you
Existing code
What the element carries today — a dash means nothing.
Object class
The classification table the code comes from. Editable.
Proposed code
The suggested code. Editable — pick a different one if the AI got it wrong.
Change
Whether this is a new code or a change to an existing one.
AI confidence
How sure the model is. Treat anything low as a prompt to check it yourself.
Apply
Untick to leave that element alone.
Accept writes the ticked proposals; Reject discards the lot.
Proposals are AI-generated (Mistral AI) and may be inaccurate — always review before use. A wrong classification silently moves emissions into the wrong part of the breakdown.
Classified elements group correctly in the project breakdown and the by-type charts. The change is recorded in History, so you can see exactly what was applied and restore the state before it if needed.
The cleanest fix is upstream: set the classification parameter in your authoring tool and map it in Settings → Classification, so the codes arrive with every publish.
