It reads your schedules, so every wall, door and window arrives typed.
Rooms detected as polygons with area, named from the drawing’s own labels. Walls as runs with length and the wall type read from the sheet’s legend or hatch: partition, exterior 2x6, 1HR rated. Doors and windows counted with their marks, and the door, window and wall-type schedules read and attached across the set, so the schedule row beats the plan label. A typed count then applies your assembly on its own. Flags on whatever it isn’t sure of, room, wall or mark.
With your review call · PDF plan sets
Architectural is the first pass in Bildrix AI takeoff software: it runs at upload, before you’ve clicked anything.
Bildrix vs Togal.AI: the full comparison →
The schedule row beats the plan label.
Door, window and wall-type schedules are read as tables and attached to the detections that carry their mark, across the whole set. A door tagged D-07 on the plan gets the size, type and rating its schedule row says, not whatever the plan label managed to fit. Walls take their type from the legend or hatch. Rooms keep the names printed on the drawing.
Press Room Schedule and the finishes join the rooms.
It reads the room legend or finish table (LÉGENDE DES PIÈCES, room finish schedule) in French or English and joins floor, wall and ceiling finishes to each room by its code. The takeoff export gains Room Name, Floor Finish, Wall Finish and Ceiling Finish columns and a Finish Summary sheet: rooms and area per finish. Each room also derives its perimeter, so wall paint prices from perimeter times the ceiling height the sheet notes, or an assumed height, labelled as assumed.
Typed counts price themselves. Untyped ones stay visibly unresolved.
Typing is what turns a detected polygon into a bid line. Each detected room, wall, door and window is typed from its name and the legend and applies the default assembly: partitions, drywall, 1HR and 2HR walls, exterior 2x6, seven door types, vinyl windows, LVP, carpet, tile, ACT ceiling, baseboard. Where nothing matches, the line stays unresolved on screen rather than priced as a partition. Where no assembly exists anywhere, the AI reads the legend, notes and schedules, writes one (studs, plates, sheathing, fasteners, labor), prices it for the job site’s market and tags it as AI work to verify.
What it does on an architectural set, and what it doesn’t.
What an estimator asks before trusting a room count, and our answers.
Does it know a rated wall from a partition?
When the sheet does. Wall type comes from the legend or the hatch, and from the wall-type schedule when there is one: partition, exterior 2x6, 1HR rated. A wall with no type on the sheet is a wall with no type in Bildrix, shown as unresolved, not guessed.
Does it read the finish schedule? In French?
Yes, both. Press Room Schedule and it reads the room legend or finish table and joins floor, wall and ceiling finishes to each room by its code. The export gains Room Name and the three finish columns, plus a Finish Summary sheet with room count and area per finish.
What happens when nothing in my library matches?
It stays unresolved where you can see it; nothing is priced as a partition by default. If nothing in any library fits, the AI drafts one from the sheet’s own legend, notes and schedules, prices it for the job site’s market and tags it as its own work. The tag stays until you replace the number.
Can it produce a Class D estimate for a Québec job?
It produces the elemental takeoff behind one. The export is a fillable CIQS Class D PDF with French element labels (CEBQ), or UniFormat II (ASTM E1557) for US jobs: Bildrix fills the quantities, you fill the rates. Walls are linear only and marked †; gross floor area is the measured room area, with an override.
Does it work in metric?
Yes. Metric plan sets are read as drawn, and the app can display areas in m², lengths in meters and amounts in Canadian dollars. The subscription is billed in USD.
How do I try it on my own plan set?
Book the review call and send a plan set: we run the first takeoff free and walk it with you. Or start in the app with 15 pages of free detection, no card.
What happens to your documents.
Bid sets are competitively sensitive.
Your plan sets stay in your workspace, stored in Amazon S3. Remove a project from your workspace whenever you choose.
Detection runs through enterprise AI APIs under terms that exclude training on your data. Your plans never train AI models.
Client view structurally hides cost and margin columns. The shared page and PDF render from it, so margin can't leak.
The signed PDF is rendered once at signing, sealed and SHA-256 fingerprinted, bound to the exact revision signed. Every later download serves those same bytes.
No SOC 2 audit yet. We're a small team and we'd rather say so than imply one. Ask us anything about data handling at support@bildrix.com.
Now onboarding founding partners.
Small estimating teams, residential GCs, and electrical contractors first. Your first takeoff is free, with your review call. Founding-cohort pricing is on the pricing sheet, with the reason it costs less than list.
See it on your plans, not ours.
The thread ends with a sealed, revision-bound estimate your client signed. Yours starts with one plan set and a call.
With your review call · PDF plan sets
