Run a drawing review
A drawing review returns a table of findings plus marked-up pages you can open in Bluebeam or any PDF tool that supports annotations.
Start in chat
- Attach the drawing set (a PDF, or a folder of PDFs) to a new conversation.
- Set thinking if this is more than a spot check. Auto (default) is Low.
- Ask for the review — for example, "Review this set for coordination conflicts and missing information."
- Answer the scoping questions. Assistant surveys the set first, then asks what to optimize for and whether you want Light, Thorough, or Comprehensive.
You do not need a workflow or a standards document to start. Assistant applies built-in discipline checklists (architectural, structural, mechanical, plumbing, electrical, civil, plus sheet-wide checks) and streams findings in as it verifies them.
For accessibility, egress, or other code questions, enable Codes Search in the same session — see Run a code compliance check.
Pick a review type
| Job | Review | Thinking | What to say |
|---|---|---|---|
| Spot check / everyday QA | Light | Low | "Light review of these sheets." |
| Full set before a send | Thorough | Low or Medium | "Thorough review of the full set." Use Medium if coordination is messy. |
| Every discipline, high-stakes QA | Comprehensive | Extra High | Pick Sonnet 4.6 Extra High, then "Comprehensive review of every discipline." |
- Light — a sample of ~5–10 sheets. ~$0.50/sheet on the sample.
- Thorough — the full in-scope set. ~$0.50/sheet.
- Comprehensive — every in-scope discipline. Same ~$0.50/sheet; more sheets and disciplines. Assistant warns and waits. Confirm only if you mean to spend.
Details and cost notes: Review types.
Add your own standards
The built-in checklists cover general constructability, coordination, and missing information. Your standards are what tell Nomic what "good" looks like on this project.
- Attach your project standards, client criteria, internal checklist, or approved example comments and ask Assistant to review against them.
- Tell it what not to flag, especially on early design packages where placeholders and unresolved details are expected.
- No formal standard yet? Start from a handful of past review comments or RFIs and ask Assistant to turn them into a checklist first.
What you get back
- A findings table — sheet, location, discipline, severity, description, and the check or standard it came from.
- Markups on the source pages — a leader line pointing at the issue with a short comment.
- A summary — counts by issue class, plus anything that was skipped.
Review the table and markups before sharing. Accept the useful findings, reject false positives, and leave unresolved items for a human reviewer when more context is needed.
Make the output more useful
- Steer with follow-ups. "Go deeper on the door schedules" or "only flag issues my GC would raise as an RFI" starts new passes without redoing covered work.
- Be concrete. "Flag any door that swings into an egress path narrower than 44 inches" beats "flag egress issues."
- Trim the standards. A few highly relevant standards work better than dumping every spec.
- Sample first. Light over a few sheets shows whether Assistant understood the task before you spend on Thorough or Comprehensive.
When to make it a workflow
Turn a review into a workflow once the same review runs repeatedly and you want every reviewer to get comparable output: instructions and standards are fixed in advance, reviewers just attach the set. For an extremely thorough, every-discipline workflow, pick Extra High thinking and write Comprehensive into the instructions. See Build a workflow from a real task.