A useful review connects a question to the documents behind it: what appears inconsistent, where it appears, why it could matter, and what needs to be verified. The output is a review aid, not a permit or a certification that the design is complete.
What does the software actually review?
Start by defining the package: the current architectural, structural, civil, and MEP drawings, plus the specifications, schedules, reports, and code references relevant to your scope. Different review tasks need different source material. A coordination check between structural and plumbing plans needs both disciplines; a drawing-to-specification check also needs the relevant specification sections.
In Groundbook, you choose which files to check, distinguish supporting references, and tailor the review instructions. The result is a list of potential findings with drawing and reference links. Your team reviews those findings in the context of the complete design.
Four useful review questions
- Do the documents agree? Compare the same level, equipment tag, material, or design requirement across the relevant sources.
- Is an interface missing? Look for an opening or connection required by one discipline but unresolved in another.
- Can the reference be followed? Check whether a detail callout or schedule reference leads to the information the reader needs.
- Does a referenced requirement apply? Investigate a potential mismatch between a drawing condition and the project’s stated code or design criteria.
These questions lead to different workflows. Explore cross-discipline coordination, drawing and specification review, and code-reference review to see their inputs and review boundaries.
What a useful finding looks like
Consider a floor elevation that differs between structural and mechanical plans. “The plans are inconsistent” is too vague. A useful finding identifies the corresponding location, points to both callouts, and asks the reviewers to confirm the datum, units, and governing revisions.
If one callout is a finished-floor elevation and the other is a structural elevation, the difference may be intentional. That context matters more than how confidently the software describes the issue. The sample floor-elevation finding shows the source pair and verification questions used to investigate this kind of result.
Where the reviewer’s judgment matters
A review can miss an issue, misunderstand a table, connect the wrong references, or apply a requirement outside its intended context. A citation is useful because it lets a person check the reasoning; it does not prove the conclusion.
The reviewer confirms that the finding describes the same component and condition in both documents, reads any relevant exceptions, and decides whether a correction is needed. For code questions, confirm the adopted edition and local amendments. For engineering questions, use the calculations and professional judgment needed to assess the design.
The ICC’s code adoption resources explain why model-code publication and local adoption are different considerations. A tool’s ability to cite a newer edition does not make that edition applicable to your project.
How to try it on a real package
- Pick a defined milestone. Use a package your team understands and a scope that can be assessed, such as equipment consistency or pre-permit coordination.
- Prepare the inputs. Include complete references, identify current revisions, and check PDF readability. Use the PDF preparation guide.
- Keep an independent review record. Record known issues before looking at the AI results so you can assess both useful findings and misses.
- Verify and classify the output. Separate actionable findings, duplicates, incorrect findings, and questions that need more information.
- Measure the work that matters. Track useful issues found, issues missed, and reviewer time spent reaching a decision. See how to evaluate accuracy.
Where it fits in the project
Use the review while the team can still act on it: before a submission, procurement decision, coordination milestone, or revised issue. A pre-permit review can help prepare the package; a constructability review focuses on document questions that could affect purchasing or field coordination.
After the designers resolve the findings, review the updated documents and dependent references. A useful first pass should lead to a better coordinated issue, with a clear record of what changed and who confirmed the decision.