The old-school way hides your numbers
The habit most of us were taught: when an estimate gets approved, print it to PDF and read it line by line to figure out what you are paying subcontractors and what the job really costs. That works, in the sense that it eventually produces an answer. It also eats hours and invites mistakes.
Xactimate will hand you the same data as a spreadsheet. Go into your token, then local, select the project, and on the right choose export to Excel instead of export to PDF. Everything after that is filtering, not reading.
One auto-sum, then filter your way to answers
First move: highlight the RCV column, auto-sum it, and change the column type to accounting so it reads as dollars. That running total at the bottom now follows every filter you apply, which is the whole trick.
Need to know what the electrician gets paid? Filter the category column down to the electrical line items and read the total. In the example walked through in the video, that is about $1,000 to the subcontractor, and a clearer view of margin. Clear that filter, go to the group code or group description column, select only the bathroom, and now you know what that one room costs.
From there the built-in table tools can pivot the same data into ACV and RCV by room type, the same instinct that turns a $359-a-month compare tool into a five-minute build. You will format some columns yourself, but the insight is instant instead of hours of line-by-line reading. The switch from PDF to Excel saved the presenter several hours per week.
Then let a model read the whole file
The video takes it one step further: paste the same exported estimate into ChatGPT and prompt it to act as a project manager. Ask for scheduling recommendations and job-level insights and it returns them immediately, sometimes surfacing things you would not have thought to check.
Treat that output the way you treat any junior analyst: useful for direction, verified before you act on it. The point is not that the model replaces your read of the job. The point is that your estimate data is only useful when it is in a form something can actually analyze, and a printed PDF is not that form.
Your estimate data is only useful in a form something can actually analyze.
