Framing Design Automation

Every job had to be detailed by hand before it could be made. Skilled detailers worked through each plan, and the queue in front of them set the pace of the whole business. Sales could win the work faster than the design team could release it.

The cost showed up in two places. Jobs sat in the design queue waiting their turn, and mistakes made under time pressure reached the factory, where fixing them costs far more than catching them at the desk.

Adding people was the obvious answer and the wrong one. Detailers with this experience take a long time to train, and the ceiling would have come back as soon as volume rose again.

Before any code was written, both sides agreed the measure: design time per job, and the number of errors reaching the factory floor. Not a satisfaction score, not a feeling about speed — the clock on a real job and the defects counted downstream.

The baseline was taken from their own records of completed jobs.

An automated detailing pipeline that takes the incoming job, applies the rules the detailers were applying in their heads, and produces the output the factory needs. The rules came out of sitting with the team and writing down what they actually do, case by case, including the exceptions.

The detailers stayed in the loop. The system does the repetitive geometry and the paperwork; a human reviews and approves before anything is released to manufacture. Anything the system is not confident about is escalated rather than guessed.

It was built into their existing workflow, so the team did not have to learn a second way of working alongside the first.

The pipeline runs on live jobs today. More jobs go through the same team, with fewer errors reaching the factory, and the design queue is no longer the thing that decides how much work the business can take.

It remains in production, and we still support and improve it.