Skip to content

How it works

From a drawing set to a cited review.

DC Copilot reads each sheet for what it's authoritative on, measures the design from the vectors, then tests it against the published standard — and shows its working on every finding.

01

Upload a vector PDF

A drawing set, not a scan. The engine works from the vector geometry — walls, dimension strings, title blocks — so it can measure, not guess.

02

Triage sorts the sheets

Each sheet is classified by its drawing name — floor plan, sanitary schedule, interior elevations, details. The proposed plan is told apart from the demolition plan, so the right sheet feeds each check.

03

Read by authority

The layout comes from the plan, the fixture counts from the schedule, the mounting heights from the elevations. Each fact is taken from the sheet that's actually authoritative on it — never blended into a guess.

04

Measure deterministically

Room sizes are derived from the vector walls and the drawing scale, validated against the dimension strings on the sheet — not from the AI's freehand estimate.

05

Test every requirement

The engine runs each rule in the requirement library against the read design and reconciles the sheets against each other, flagging where they disagree.

06

A cited report

Findings come back grouped — must-fix, meet-where-practicable, review, pass — each with the verbatim clause and reference. You confirm each one and decide.

Read by authority

Each fact from the sheet that owns it.

A single AI pass over a whole set blends everything into one stochastic answer. DC Copilot instead takes each fact from the sheet that is authoritative on it, then reconciles them — so a count comes from the schedule, a height from the elevation, and a size from the measured plan.

Floor planRoom layout, fixtures, clearances
Sanitary scheduleFixture counts (pans, basins, urinals, showers)
Interior elevationsMounting heights (basins, rails, mirrors)
Finishes scheduleMaterials, coved upstands, slip resistance
ReconcilerCross-checks the sheets and flags disagreements

Citation-first

The AI reads. The engine rules.

The language model's only job is to read the drawings. The compliance logic is deterministic code that tests each requirement — so a finding is never an AI opinion. Every one is pinned to a verbatim clause and reference you can check against the source in seconds.

Stage-aware

Checked to the stage you're at.

Concept, preliminary or developed — each requirement knows when it can fairly be assessed. You won't be failed for detail you haven't drawn yet, and you'll see what still needs to be shown as the design develops.

See it on a real set.

Run the built-in sample, or upload your own drawing set.

How it works · DC Copilot