Mapping the process before automating it, for a client seven states wide

Construction estimating across my own companies traditionally took about six weeks per job. Separately, a multi state cannabis operator needed unified standard operating procedures across seven states, each with its own regulatory complexity. Instead of buying off the shelf software for either problem, I built proprietary automation for plan scanning, takeoffs, and market specific cost correlation, then applied the same method, map the process first, then automate it, to standardize the cannabis client's SOPs across all seven states. I treat AI as engineering for a workflow I already understand, not a shortcut around understanding the business. Modeling the process first means what gets automated is grounded in metrics I already trust, which is exactly why the same method transferred cleanly across two completely different industries. Estimating that used to take six weeks now takes about six hours internally, and the cannabis client got standardized SOPs across seven states without manually reconciling each state's rules by hand. Automation only works if you already understand the process well enough to know what's actually worth encoding. I didn't sell either client a black box. I built the automation on top of a process I'd already proven I understood by hand.