Coordination errors hide in the boring layer of a drawing set: a callout pointing to a detail that does not exist, a revision out of step between sheets, a missing legend. They are cheap to fix before issue and expensive after it, but the pre-issue check is manual, repetitive, and usually squeezed against the deadline.
A checker that reads drawing metadata and extracted notes, runs a fixed set of rules over them, and returns a findings list grouped by severity. The rules cover missing titles, duplicate sheet numbers, revision mismatches, callouts to missing sheets or details, details that exist but are never referenced, and missing legends or general notes.
Load a sample drawing set, work through the findings, and mark each one reviewed. When the pass is done, the open findings export as a QA register, either the full list or just what is still outstanding. The sample sets include seeded errors, so you can judge for yourself whether the checker catches what it claims to.
Every finding carries its sheet reference and a suggested review action, so it can be checked against the set rather than taken on trust. The reviewer closes findings; the tool never closes its own. What the checker cannot see, like design intent or dimensional correctness, stays out of scope and is said to be out of scope.
Whether a lightweight second-pass reviewer earns a place in the issue workflow. The bar is practical: it has to find real omissions faster than a person paging through the set, without flooding the reviewer with noise, and without anyone mistaking it for the QA itself.