Newsletter
The Rejection That Costs You Three Weeks Is Almost Never an Engineering Mistake
A missing sheet or wrong permit citation can cost three weeks. How MES split set completeness checks from design judgment, and where the tool's limits are.
Modern Engineering Solutions
Engineer Your Freedom
The Modern PE
September 2026
The rejection that costs you three weeks is almost never an engineering mistake.
Half of plan review needs a license. The other half needs patience nobody has at 4:45 on submittal day.
A set comes back from the agency without a single technical comment.
The index lists 77 sheets and the file has 69. Or a note on a Texas project cites a Colorado permit number. Three weeks of calendar are gone and the reviewer never reached the design.
77
Sheets listed
in the index
−8
Sheets short
69
Sheets actually
in the file
Reviewing a plan set is two jobs we have always treated as one. The first is design judgment. Is the pipe sized right, does the detention approach hold. That needs a licensed engineer and always will. The second is reading the set against itself. Does every sheet in the index exist in the file. Does every detail callout land on a real detail. Is one sheet still stamped 60 percent in an issued-for-bid set.
The second job requires no engineering judgment at all. It requires patience, and patience is the first thing compressed to nothing when a deadline lands. It also causes most of the rejections that come back with no review attached.
Two Ways A Set Fails Small sets
Fail by inheritance
Built from the last project's file
- Previous inspector's name and phone still in the notes
- A template placeholder reading CITY that shipped to the agency
- Proposed work still described as existing because the base was an as-built
Fail in the gaps between sheets
Errors that never show up one page at a time
- Sheets listed in the index and absent from the file
- Detail callouts pointing at sheets that lost that detail in a late renumber
- Two vertical datums in one set, or a broken match line
A careful person catches the small-set errors in twenty minutes, and nobody has twenty minutes on submittal day. On an 80-sheet set, a person reads the sheets one at a time, and none of the gap errors show up one at a time.
The expensive errors on a large set are the ones that live between the sheets.
A reviewer holds one page at a time. Software holds the whole set at once, and that is the entire advantage.
Most Of It Is Not AI
That is why I let it run ahead of a licensed engineer. Set completeness, slopes, structure depths, curb heights, and ADA limits are all computed in code. The model reads drawings into structured data. Of 216 checks across 14 sections, the ones that must never be wrong never touch a language model.
216
Checks across
the full set
14
Sections in the
master checklist
80
Sheets in the
example large set
<1 hr
PE time cost of a
full large-set pass
Worth telling you where it failed too. An earlier version returned 138 real findings on a large treatment plant set and missed that eight indexed sheets were absent. Volume of findings is not coverage. That miss drove the completeness work.
The Boundary
What the tool can and cannot see
It reads
- The text layer of the PDF
- Drawings parsed into structured data
- Set completeness against the index
It cannot see
- A seal that exists only as a graphic
- A water line drawn with the wrong symbol
- A pure raster sheet, which goes unreviewed
When a sheet goes unreviewed, the report names those sheets rather than staying quiet. Teaching it to read the drawing itself is the next build, and the subject of the next letter.
What changes is where your people spend attention. Your PE stops being first eyes on whether sheet 47 made it into the file and becomes first eyes on whether the design is right. Your EIT stops learning review by proofreading and starts learning it by watching judgment applied to real problems.
The cost of a review run is not a decision variable.
A full pass on a large set costs less than one hour of a licensed engineer's time, and it is not close. Whether to run it was never the question.
The Honest Limit
I will not give you a catch rate, because we have not measured one. The arithmetic is verifiable by construction. A slope matches the invert math or it does not. The judgment-level findings are not characterized yet, and I would rather say so than sell you a percentage.
The Question I Want Your Answer To
You bill quality control today. It sits on a timesheet, or it is buried in a lump sum fee your team priced around a PE spending an afternoon on it. When that afternoon stops being an afternoon, what happens to the invoice?
On hourly
You deleted revenue
On lump sum
You created margin
Most firms have not decided which one they are.
Reply and tell me how you would charge it.
That answer says more about your delivery model than any software decision you will make this year.
Michael Groselle, P.E.
Founder and CEO, Modern Engineering Solutions
mod-eng.com | contact an engineer | 214-833-6748
Talk to an engineer
Tell us about your project.
An engineer reads it
Not a salesperson. Someone who can tell you whether the pathway is real.
You get a first answer
Your options, the agencies involved and the questions that decide cost and schedule.
You decide
No commitment to have the first call.
Prefer the phone? (214) 833-6748