Your Detailer Retypes 62 Handwritten Door Widths Into Cabinet Vision. That Is Where the Remakes Come From.

It is 4:15 on a Thursday afternoon and your lead installer is standing in a half-finished stairwell on the third floor of a medical office building, holding a clipboard, a laser measure and a carpenter's pencil. He has sixty-two openings to field-measure before the general contractor locks the site at five. He writes each rough opening width in sixteenths, then the height, then the jamb depth, then the hand, then a note in the margin that says "hold 1/4, carpet not in yet." He photographs all four sheets against the drywall, texts them to the shop, and drives home.

By eight the next morning somebody in your detailing room has retyped every one of those numbers. That retyping is where your remakes come from. Not the moulder, not the finish room, not the installer with the pencil. The keying.

The field-dimension sheet is a manufacturing document that everybody treats like a note

Look at what is actually on that sheet. Opening number, tied back to the door schedule in the architect's Division 08 14 00 pages. Rough opening width and height in sixteenths. Jamb depth, which on one job is 4-9/16 and on the next is 6-9/16 and on a CMU corridor is whatever the mason left you. Hand and swing, written as LHR or RHR by a man wearing gloves. Hinge count. Whether the opening is labeled, because a labeled opening changes the whole order. Then the margin: "wall out 3/8 at head," "return in 3 weeks, CMU not grouted," "GC says hold this one."

That sheet drives a moulder setup, a door machine setup, a jamb cut, a stain match and a delivery ticket. It is worth several thousand dollars of material and finished labor. And in most millwork shops in this country it makes the trip from the job site to the shop floor by way of a person squinting at a phone photograph and typing into Cabinet Vision, Microvellum, Mozaik or a spreadsheet that feeds the CNC.

Multimodal document reading means software that takes the photograph, the fax or the scan exactly as it arrives, creased and initialed and shot at an angle in a stairwell, and pulls the numbers off it, instead of requiring a person to retype them into a clean form. That is the thing that got genuinely good in 2026, and it is the reason this post exists.

Sixteenths, sevens that look like ones, and the 0.6 percent nobody measures

Nobody at your company tracks the keying error rate on field measures, because the error does not appear as a typing mistake. It appears three weeks later as a door that is 3/16 narrow, a jamb cut for the wrong wall thickness, or a stile-and-rail unit machined for the wrong hand. By then it is a remake, a second finish cycle and a second trip.

The handwriting is the whole problem. A hurried 7 reads as a 1. A 5/16 reads as a 5/8 when the pencil is dull. A circled note gets skipped because the detailer is working from a photograph where the flash washed out the top right corner. And the sheets do not come in one format either: your commercial installer uses your printed field-dimension form, the residential crew uses a legal pad, and one builder still faxes you a marked-up door schedule with initials in the margin.

What the reading step looks like when nobody retypes it

flowchart TD
  A["Installer field-measures 62 openings"] --> B["Photo of the field-dimension sheet texted to the shop"]
  B --> C["Reader pulls opening no., RO width, height, jamb depth, hand"]
  C --> D{"Does every line match the door schedule?"}
  D -->|No| E["Flagged lines go back to the installer before anything is cut"]
  D -->|Yes| F["Cut list written into Cabinet Vision"]
  E --> F
  F --> G["Shop foreman releases to the moulder and door machine"]

The difference between this and the 2024 version of the same promise is that the old readers wanted a clean scan of a form they had been set up for in advance. Hand them a crumpled sheet photographed on a job site and they returned garbage or nothing. The 2026 readers handle the ugly input: the shadow across the page, the correction written over the original number, the arrow pointing from a margin note to a specific line, the initials next to a change. They read the artifact the way your detailer reads it.

Tuesday morning in the detailing room, the new way

Your detailer opens the text thread. The four photographs are already read. On her screen is a table of sixty-two lines: opening number, width, height, jamb depth, hand, label, margin note carried across as text next to the line it belongs to. Eleven lines are highlighted. Nine because the field width differs from the door schedule by more than an eighth. One because the hand written on the sheet contradicts the swing shown on the architect's plan. One because the reader could not resolve whether the number is 5/16 or 5/8 and says so plainly instead of guessing.

She spends twenty minutes on eleven lines instead of ninety minutes on sixty-two. She texts the installer two photographs of the specific lines in question. He answers from his truck. She releases the cut list. The moulder runs Wednesday.

The second thing worth having: the original photograph stays attached to the job. When a superintendent calls in September claiming the opening was always 3'-0" and you shipped 2'-10", you open the job and there is the sheet, in the installer's handwriting, with the date on it.

What one remake actually costs, and the arithmetic on catching it

Assumptions, stated plainly, and they are illustrations rather than measured facts. Suppose your shop details 900 openings a month across commercial and residential work. Suppose the keying error rate is 0.6 percent, which is roughly one bad line in every 170 typed. That is five wrong openings a month.

Against that, the reading itself is close to free at your volume. Running this on every field-measure sheet your shop receives costs less than what you spend on shop rags. The real cost is the two weeks somebody spends teaching it your form layout, your abbreviations and your opening-number convention, plus the discipline of making installers photograph the whole sheet instead of the top half.

Where this still needs a human, and it is not a small list

It does not decide anything. If the field width and the door schedule disagree, the reader flags it; a detailer or a project manager still has to call and find out whether the wall is wrong or the schedule is wrong. That judgment is your business.

It will not read a dimension the installer never wrote. A blank is a blank. If your field-measure discipline is poor, this makes the poor discipline visible faster, which is useful but is not the same thing as fixing it.

Labeled openings deserve extra caution. On a fire-rated assembly, the manufacturer's listing governs what can be prepped and what cannot. Let the reader capture the numbers and let a human confirm the label line every time. The same goes for AWI Quality Certification Program jobs, where the certified label depends on the shop drawings matching what actually shipped.

And it will not settle a backcharge argument on its own. It gives you the dated photograph and the extracted line. A person still negotiates.

The Monday version of starting

Do not begin with sixty-two openings on a live job. Pull twenty field-measure sheets from jobs you already delivered, ones where you know the outcome. Have the reader pull them. Compare its table against what actually got built. You will learn two things in an afternoon: how often your own sheets are ambiguous, and whether the reader catches the specific handwriting habits of the two installers who do most of your measuring.

Then run it in parallel for a month. The detailer keys as normal, and the reader's table sits next to hers. Count the disagreements. When the disagreements are consistently the reader catching her, and not the other way around, stop keying.

Frequently asked questions

Our installers measure on paper and always will. Does this require them to change?

No, and that is the point. The paper sheet on the clipboard stays. What changes is that the photograph is the input rather than a reference the detailer squints at while typing. If anything, ask installers for one small change: photograph the sheet flat, whole, with the opening numbers visible, rather than four close-ups.

We still get faxed purchase orders from two builders. Can it read those?

Yes. A faxed order with a handwritten quantity change and initials in the margin is exactly the input the 2026 readers were built for. The margin note is the part that used to be lost, because the old readers pulled the printed fields and ignored anything written on top of them.

Will it write straight into Cabinet Vision or Microvellum for us?

Sometimes, depending on how your shop imports cut lists, but do not make that the test. The saving is in not retyping and in the cross-check against the door schedule. Even if the output lands as a reviewed table your detailer imports, you have taken the error out.

What about the specification and the submittal package, not just the measures?

Same reading, bigger document. A marked-up door and hardware schedule, an ASI, a revised finish schedule with a wet stamp on it, all read the same way. Many shops start there, because a missed revision on a submittal is more expensive than a missed sixteenth.

One more place the same reading shows up: the phone. Most field-measure corrections arrive as a call to the shop at 7:20 in the morning, when the detailer is not at her desk yet. CallSphere builds AI voice and chat agents that answer the line, take the caller through what changed on which opening number, and drop the note onto the job before anyone starts cutting. It does not read your sheets or run your machines. It makes sure the call about the sheet does not get lost.