Estimators stopped keeping secrets, and that changed everything
The video starts with a post in Restoration Rebels, a Facebook group where much of the industry congregates: what line items are we leaving off the table? A generation ago estimators guarded their code knowledge. Now those questions get asked and answered in public, in group after group, year after year.
That archive turned out to be training data. The video walks through searching for Xactimate line item discussions, extracting the threads, cleaning them, and training a model on the result alongside the builder's own estimate history. The model now carries not just one estimator's habits but years of the industry's collected questions and answers.
Photos in, common line items out
The demo: two photos of a kitchen hit by a water loss from upstairs, piped into the system. On the back end it searches the builder's written estimates and their photos, finds losses that look like this one, and returns the line items most commonly used on those similar losses.
The claim is not that the output is a finished estimate, though that is where this work eventually went. It is that a large chunk of the estimate is already drafted before you look at a single photo yourself. The estimator starts from a list grounded in how comparable losses were actually scoped, then confirms, corrects, and completes it.
What this buys the desk
Two things, per the video: more detailed scopes, because the model surfaces codes you might have left off the table, and more efficiency, because the first pass is written for you. Time saved at the desk is time back in the field.
This early work is the foundation of what Axiom does today. Photos and job details build a room-organized draft with suggested codes and quantities, and your estimator reviews all of it before downloading the ESX. The point was never to replace the estimator. It was to stop making them start from zero.
The point was never to replace the estimator; it was to stop making them start from zero.
